At the Maddox farmhouse, no one was allowed to sit alone. A chair would be found, a place made, a little more water added to the soup; Elizabeth Maddox could accommodate anyone, and for thirty years her friends mistook this for a house rule. There were people who came for an evening and remained until their divorces were final. There were children who arrived with their parents and returned without them. Outside the kitchen window stood an oak that had grown around an iron fence post, a length of washing line, and the hook from a swing. What it could not move, it made room for.

The oak stood for a hundred years, and for a hundred years it was asked to hold what the house could not: a rope swing frayed where small hands had held on, initials carved lovingly and not very well by some boy in love, the weight of children climbing up its outstretched arms – and, in the garden, the other end of every conversation Elizabeth Maddox ever had with herself. It held them all. Holding is the only thing a tree knows how to do, and it does not distinguish between a tired swing and a hushed secret. Three grown people holding hands could not reach around it, and its bark was furrowed deep enough to lose a finger in. On the side toward the house a limb had come away long ago, and the bark had closed over the place like a mouth; apart from that, it had never been known to drop anything it held. Every afternoon in late summer its long shadow came in through the kitchen window, took a seat at the table, and stayed until after dark.

When Elizabeth died, the farmhouse was sold. The new owners wanted light in the kitchen, and the oak came down on a Thursday in the autumn of 2019. It took four men most of the day and three saw chains. The blade found the fence post first, then the washing line, each of them deeper in the trunk than a hundred years of growing could have carried it. It found the hook from the swing, though the limb the swing had hung from was long gone. Deepest of all, at the very heart, it went through the sole of a child’s boot. The man holding the saw looked at it for some time and did not call the others over. They cut the trunk into lengths, ate their sandwiches, and left. After dark, someone came back for a piece.

It was Will Brandt, who had once lived in Elizabeth’s spare room for six weeks while he found somewhere else to be unhappy. Since her funeral he had driven down from Olympia most Thursdays and parked across the road, where he could see the kitchen window, and he had watched the whole day’s work from his car. He used a handsaw. Halfway through, the cut closed on the blade, and he had to put his foot against the trunk and pull. He fell backward when it came free.

There was nothing remarkable about this. Green wood can pinch a saw. Nevertheless, he apologized.

The trunk had been cold wherever he put his hand. The length he carried to the car was warm, and it was still warm on the seat beside him when he reached Olympia.

* * *

Seven years later, Claire Maddox invited four people to a cabin on Orcas Island. She wrote the invitations by hand. First weekend of November. I need all of you there. Please don’t make me explain in a letter.

Diane Weiss called her immediately. Claire let it go to voicemail, then sent a photograph of her ferry reservation.

“That manipulative little shit,” Diane told Phoebe, and booked a ticket.

She arrived with enough groceries for a week. Marcus Chen was already there, aligning five chairs around a table intended for four. He had found the fifth in a bedroom. It was a straight-backed kitchen chair with a towel folded over the seat, and he carried it out with the towel still on it.

“Where’s Claire?”

“Lying down.”

Something in the way he said it made Diane leave the groceries on the floor.

Claire was awake. She lay on top of the bedspread in a cardigan Diane had given her ten years earlier. The buttons were little wooden fish. One was missing, and the cardigan hung open over the place where there had been a body with some weight to it.

Diane sat beside her.

“How long have you known?”

“Since August.”

“You could have called.”

“You’d have come.”

“Yes, Claire. That’s generally the fucking idea.”

Claire took her hand. Her wedding ring was on a chain around her neck, although the marriage had been over for eleven years. She wore it when she needed something to worry between her fingers.

“I wanted one dinner before we did this.”

Diane looked at the little fish. She could sew a button on. She had needles in the car, and the glasses she needed now to thread one, in the kit Phoebe made her carry because she insisted on doing everything herself.

“All right,” she said. “One dinner.”

In the kitchen Marcus was putting away the food. He practiced family law in Tacoma and had become very good at knowing when to leave a woman alone. He passed Diane a mug of whiskey.

“Phoebe says hello,” she told him.

“How’s Chloe?”

She looked at him.

“Jesus. Phoebe. I’m sorry.”

“She’s fine. The alpacas are fine. Everyone you haven’t called is fine.”

He drank from his mug. “I’ve missed you too.”

That was enough to let them begin.

They chopped onions and disagreed about how much garlic Claire could tolerate. He told her she had gone gray. She told him he hadn’t, and that at fifty-seven it was a decision. Rain moved across the channel. When Diane reached past Marcus for a towel, he put his hand on her shoulder for a moment, and she let him. She had expected to be angrier with him. There was always less time in a room with the actual person than there had been in the arguments she held alone.

The soup was Elizabeth’s, the one she could stretch for however many came. Her kitchen had always smelled of it, and under the soup there had been another smell, of wet earth and cut stems, because she was forever trimming something at the sink. There had been one chair at that table nobody ever saw empty. If you got up from it, Elizabeth sat down in it herself and went on talking. Anyone left standing was told, “Sit down, sweetheart. You’re hovering.”

Tom Aldiss arrived at dusk. He came in carrying a paper bag, gave it to Diane, and went back for his coat.

Inside were five satsumas. “He stopped at a store,” she said, very quietly.

Marcus nodded.

For seven years Tom had done only what needed doing. He ate, worked a little, kept his truck running. He had told Marcus once that he kept one chair at his table now, since one was what got used. At Christmas he answered their messages with the same three words: You too. Thanks. His son Ethan had died at nineteen, in February 2019, and there were parts of ordinary life Tom still approached as though someone had moved the stairs in the dark.

But he had stopped at a store. He had counted them.

Will Brandt came last. His hair had gone completely white. He stood in the doorway with a canvas bag over his shoulder until Claire emerged from the bedroom and said his name. He put the bag down so carefully that Diane looked at it twice.

At dinner they did better than she had hoped. Marcus told a story about two parents who had hired competing psychics in a custody dispute. Will asked which one had been right.

“The child wanted to live with his grandmother.”

“Were either of them getting that?”

“No. I was getting that.”

Tom smiled into his plate. The lines came up beside his eyes, and Diane realized she had forgotten he had them.

They ate Elizabeth’s soup, which tasted the way it always had, a little thin and exactly right, and for an hour they were very nearly the people who used to eat it.

They had met at Reed, except for Tom, who had arrived attached to an engineering project and stayed attached to them. Most of the friendship had happened in Elizabeth’s kitchen outside Portland. Elizabeth fed them and let them be clever until they became unkind. Then she would lay her hand on the table and say, “Tread softly,” and whoever was winning would discover a sudden interest in the salt.

It had taken Diane years to learn that the words came from a poem. Elizabeth had seemed quite capable of inventing them for the occasion.

After dinner Claire gathered the plates, got as far as stacking two, and sat down again.

“I have pancreatic cancer,” she said. “They can treat it. They can’t fix it.”

Marcus reached for his mug and left his hand beside it.

“How long?” Tom asked.

“Months. They’re vague. I used to find that annoying in a man.”

Nobody helped her with the joke.

“Look,” she said. “You’re all going to be wonderful about it, and I’m grateful. But there’s something I want to do while I can still get on a ferry by myself.”

“What?” Tom asked.

“I want to ask my mother something. One thing.” She turned the ring on its chain. “Whether she’ll be there.”

She looked at Will.

He did not move.

“You promised.”

“I promised to bring it.”

“Then let them see.”

* * *

The board was smaller than Diane expected. She had pictured something expensive and ridiculous, an object with brass fittings. Will took out a shallow oak frame with a sheet of cream linen stretched across it. A smell came out of the bag with it, very faint, of wet earth and cut stems. Diane knew it at once and did not say so. He had painted the alphabet in two rows, numbers below, YES in one corner and NO in the other. GOODBYE ran along the bottom.

The pointer was a wooden triangle with a hole through its center. The grain curved around the hole like the lines on the pad of a finger. In one corner Will had blotted the ink with his thumb and painted over the mark. She could see his worktable in Olympia, the newspaper laid down, the small brush between his fingers. This had taken him evenings. He had gone to bed thinking about it and got up wanting to finish.

“A Ouija board,” Marcus said.

“A talking board.”

“Oh, well.”

“He made it from the oak,” Claire said.

Diane looked up. “Your mother’s?”

Will nodded.

“I thought if anything of her was left, it would be in that tree,” he said. “It was on that farm seventy years before she was. I didn’t think about that part until later.”

“And she talks to you on it.” Marcus waited. “Is that what we’re being asked to believe?”

“I’m asking you to sit with me,” Claire said.

“Those are different things.”

“Then do the easier one.”

Diane cleared the plates. She was annoyed by the candles Will requested and more annoyed by the part of herself relieved when he said the electric lights could stay on. Outside the window the channel was black. A little kitchen clock ticked over the refrigerator.

They sat where they had eaten, Claire at the end, Tom and Marcus along one side, Will and Diane along the other. Diane had the chair from the bedroom. Without its towel, the seat was cold through her jeans.

Will put two fingers on the triangle.

“Very lightly. It’s called a planchette.”

“Will,” Diane said. “Before we do this. What happened when you used it?”

He kept his fingers where they were.

“Usually nothing. Sometimes answers.”

“From Elizabeth.”

“That’s what they said.”

The kitchen clock made eight small ticks.

“Once it called me Billy,” he said. “She never called me that. No one does.”

“You’re admitting you might have imagined it,” Marcus said.

“That would be preferable.”

“And you went on using it,” Diane said.

“Nothing else in my house answers.”

Claire was watching the board.

“Have you used it?” Diane asked her.

“With Will. Twice.”

“And?”

“The second time it told me to bring you.”

“All of us?”

“By name.”

Diane sat back. She was conscious of the cold seat, of the bottle on the counter, of Tom’s hands flat on his knees.

“Claire, I don’t think this is a good idea.”

“Neither do I.” Claire’s voice was level. “But I’ve had months of people telling me what they don’t think is a good idea. I’d like an evening off.”

Will lifted the triangle and showed them its underside. Three felt feet, a shallow crack in the wood.

“There’s one thing. If it starts moving through the numbers backward, take your hands off. Don’t wait for me.”

“Why?” Tom asked.

“The last time it did that, I stopped at six. I put it away in a pillowcase and tied a knot in it. In the morning the knot was still there. The pointer was on the stairs.”

“You should have led with that,” Marcus said.

Will put it down.

Nobody reached for it. Rain ticked against the window, briefly louder than the clock. Diane could get up. She could take Claire into the bedroom, close the door, sit on the bed again and explain. She had spent thirty years as a veterinarian telling people things they did not want to hear. She knew how to keep her voice steady.

Claire laid her fingers on the wood.

Tom followed.

Diane looked at his hands, the wide pale band where he no longer wore his wedding ring. She put two fingers beside his. Across the table Marcus released a breath through his nose and joined them.

“Elizabeth,” Will said. “We’re here.”

Nothing happened.

The cramped arrangement of their fingers made Diane think of a game Ethan used to make them play, each person adding a hand to a pile until the hand at the bottom had to come out and slap the top. Elizabeth always lost. She would leave her hand underneath, apparently forgetting what they were doing, and Ethan would dissolve into his peculiar little wheeze before he could get a laugh out. He accused her of losing on purpose. She denied it so badly he laughed harder.

“Is anyone here?” Will asked.

The planchette turned beneath their fingers.

Diane looked at Marcus. He was looking at her.

It stopped on YES.

“Did you do that?” she asked.

“No.”

She could feel a tremor through the wood. With so many hands on one object, it could have belonged to anyone.

“Who is this?” Will said.

C-L-A-I-R-E.

Claire frowned.

“Mom?”

YES.

“Why wouldn’t she say Elizabeth?” Diane asked.

The planchette started moving again before Will could answer.

YOU BROUGHT THEM.

“I said I would.” Claire’s chin had begun to tremble. “I’m here. We’re all here.”

GOOD.

Diane looked at Will. His eyes were wet.

“Ask it something,” Marcus said. “Something only she would know.”

Claire wiped her cheek with her free hand.

“What did you sing when you washed my hair?”

The wood pressed against Diane’s fingers, paused, moved.

YOU NEVER LIKED MY SINGING.

Claire made a small, broken laugh.

“No.” She put her knuckles against her mouth. “No, I didn’t.”

Marcus laughed too, in spite of himself, and Tom’s mouth had gone soft. For a minute it was the old table, five people leaning in over something that was going well, and Diane felt the cold seat under her and did not mind it.

It was a lovely answer. Diane disliked it at once. You could give it to almost anyone. Yet Claire had begun to laugh in earnest now, with her free hand over her eyes, and Diane remembered Elizabeth at the sink with a towel over one shoulder, making a noise that rose without warning above their conversation. Someone would protest. Elizabeth would get louder. It had irritated her then. She would have given a great deal to stand in that kitchen and be irritated by it again.

“Ask for the song,” she said.

But the planchette was already moving, and Claire was following the letters with her whole body, leaning forward as if she could get nearer to the hand that wrote them.

WE NEED TO TALK ABOUT THE BOY.

Tom took his hands away.

“No,” he said.

The triangle kept moving. Diane felt his absence as a sudden imbalance, a cup tilting toward her.

HE TOLD ME.

“Stop.” Tom stood. His chair struck the wall behind him.

Will lifted his fingers, and Diane did too. Claire held on another moment, then let go. The pointer stopped between two letters.

“I didn’t ask about Ethan,” Claire said.

Tom was looking at Will.

“I wouldn’t,” Will said.

Marcus picked up the board. He turned it over, nearly upsetting a candle, and ran his hand across the back. There was nothing to see but linen, staples, the penciled numbers Will had used to match the corners of the frame.

“Give it to me,” Claire said.

“Enough.”

“It’s my mother’s tree.”

“It’s a piece of wood.”

“Then you shouldn’t be frightened of it.”

Tom went to the front door. Diane got up to follow him, but he held out a hand without turning. The latch worked. Cold air came in, and with it the sound of rain in the trees. He stood in the open doorway for a long time.

Diane could see his truck. He could reach it in ten seconds.

“What did he tell her?” he said.

“Tom.”

“You heard it.” He looked at Diane now. “‘He told me.’ What did he tell her?”

She knew precisely what her face had done. She felt it happen and could not prevent it.

Tom shut the door. He kept his hand on the latch, then let it fall.

* * *

They sat down again without discussing it. Marcus replaced the board. This time Tom reached for the planchette first.

“What did my son tell you?”

The pointer moved so slowly they had time to form each possible word and abandon it.

HE DIDNT WANT TO LIVE.

Tom stared at the hole in the wood.

“When?”

LABOR DAY.

“That last weekend at Mom’s,” Claire said. “I came on Sunday. Was he there before I got there?”

“Saturday,” Tom said. “He drove up with me.”

“Mom never told me.”

SHE TOLD DIANE.

No one turned abruptly. That was worse. They looked at her with an awful care.

“She called on Monday,” Diane said.

Tom did not move.

“She said he’d talked to her in the garden. That he was frightened of himself. She wanted me to talk to him.”

“Did you?”

“I gave her the name of someone in Portland. A psychiatrist.”

“Did you talk to Ethan?”

“No.”

The kitchen clock ticked. Diane wanted a drink of water. The glass stood six inches from her free hand, and she could not make herself reach for it.

“Why you?” Claire asked.

“Because in 2004 I tried to kill myself.”

Will’s fingers tightened against hers.

“Elizabeth drove me home from the hospital. I stayed with her for a while.” Diane looked at the candle in front of her. There was a gnat in the wax, intact beneath the surface. “She thought I would know what to say to him.”

“You might have,” Tom said.

“Yes.”

“You might have tried.”

“Yes.”

“Instead you gave her a number.”

She had rehearsed this conversation for seven years. In the rehearsals she managed to explain what it had cost to be well again; how carefully she had arranged her days, how frightened she was of disturbing anything. Sitting across from Tom, she could remember only that Elizabeth had called during breakfast. Diane had put her on speaker and buttered a second piece of toast.

“I told her to tell you,” she said. “She was afraid he’d deny it. That he wouldn’t trust her again. I said to give him a few days to make the appointment himself.”

“And after a few days?”

“I didn’t ask.”

Tom’s eyes had gone bright.

“Did it help you?” he asked. “Having somebody come and get you?”

Diane lowered her head.

“Look at me.”

She did.

“Did it help?”

“Yes.”

Under the table something touched her ankle.

She jerked her foot back. Will glanced at her. Both his feet were visible, drawn beneath his chair.

The touch came again, higher this time, against the outside of her calf: a brief, clumsy pressure. At Elizabeth’s table, someone had always reached a foot across to tell the person being scolded that they were still loved.

Diane pulled both feet onto the rung of her chair.

Tom was still looking at her.

“I should have called,” she said.

The pointer moved.

MARCUS HAS THE LETTER.

Marcus’s hand came off so quickly he caught Will’s knuckle with his nail.

“What letter?” Tom said.

“I need some air.”

“What letter, Marcus?”

He was already standing. Diane looked down at the blood gathering on Will’s hand, a single red bead beside the nail. Will seemed not to have noticed.

“Elizabeth wrote to you after Ethan died,” Marcus said. “She gave it to me when I visited her in April. She asked me to deliver it.”

“You didn’t.”

“She was very ill.”

“I know how ill she was. I went to her funeral.”

“I thought you had enough.”

“You thought.”

“Yes.”

“Is that all it took?”

Marcus looked toward the door, then sat without bringing his chair closer. He had left a gap between himself and the table.

The planchette turned.

GREEN BOX.

“Oh,” Claire said.

For a moment she only looked at him. Then she touched the ring hanging from her neck.

“My mother’s emerald. You said you hadn’t seen it.”

Marcus folded his hands in his lap.

“It was on her dresser when we cleared the house. The box was open. I put it in my coat.”

“Why?”

“I don’t know.”

“You do.”

He looked at her then.

“I wanted something.”

Claire released the ring on its chain. It clicked against a wooden button.

“So did I,” she said.

Diane heard the little sound clearly. She could hear the clock, the rain, Will breathing through his mouth. Something was missing. She looked at the window, where darkness pressed flat against the glass, and tried to think what it was. Then she knew: the water. Since she arrived there had been a soft drag and collapse of waves against the shore. She had stopped noticing it, but now its absence made every sound inside the cabin seem too close.

“And the letter?” Tom said.

“Under the tray. In the box.”

“Where?”

“My car.”

“You brought it?”

Marcus looked at Claire. “I was going to give it back.”

“When?”

Nobody rescued him from the question.

“Go and get it,” Tom said.

Marcus stood. This time no one objected. He took his keys from the counter, opened the door, and went out into the rain. Before he pulled the door shut there was the rain, and still nothing from the shore.

Diane watched him through the window. His car lit up. He bent into the passenger side. For several seconds she thought he would get in and drive away, and she wanted him to. She wanted somebody to do the obvious thing.

He came back with water in his hair and a small green box in his fist.

“Hands off,” Diane said.

They took their hands off. She picked up the planchette and set it on the kitchen counter.

The ring lay in green velvet. Marcus lifted out the tray. Beneath it was a folded sheet of thin blue paper, split at one crease from having been opened many times.

Tom reached for it.

Marcus did not immediately let go.

“There are things in here she says about all of us.”

“It has my name on it.”

Marcus released the paper.

Tom read standing up. Diane could see Elizabeth’s handwriting through the sheet, its long, impatient tails. An e was always nearly an i. Even from the wrong side she knew it.

Tom reached the bottom and started again at the top.

“Why didn’t you give me this?”

“I told you.”

Tom turned the page outward. He put a finger beneath two lines halfway down.

“Read it.”

Marcus did not.

“Read what she says about you.”

“Tom, please.”

“I’ve read what she says about me.”

Marcus looked at the paper.

“‘I am sending this with Marcus because he knows what I am saying is true. Whether he gives it to you will depend on whether he can bear to be disliked.’”

No one spoke.

“She put it in an envelope,” Marcus said. “I opened it in the car.”

“You could have put it back.”

“I know what I could have done.” He raised his voice, and it sounded so much like the old Marcus, the one who could be funny and contemptuous in the same breath, that Diane almost expected Elizabeth to intervene. “She gave me a sealed letter and arranged for you to know I’d read it. She knew exactly what I’d do. She put that in there and she knew.”

“Yes,” Claire said. “That sounds like her.”

Marcus looked at her. She was crying now, quietly, and smiling at something he could not see.

Diane remembered Elizabeth at her most difficult: the precise remark, the eye that followed you after she made it. How eagerly they had joined in when it was somebody else. Once, over breakfast, Elizabeth had described the kind of woman Marcus would marry. They had laughed until he left the table. Later Diane had found him sitting in his car. She had knocked on the window and said, “Oh, come on,” because she wanted him to come back inside and make it all right. She could not now remember what Elizabeth had said, only how certain she had been that Marcus ought to take it well.

Tom laid the letter on the table. The page had gone soft where his hands had held it.

“She says he talked to her beside the oak. She says she meant to tell me before we left.”

His voice had changed. The anger was still there, but it no longer knew where to go.

“She says I was happy that morning. That he was making me laugh. She didn’t want to spoil it.”

Claire put a hand over her mouth.

Tom pulled out his chair. He sat with the paper between his forearms.

“We’d had a good drive up,” he said. “At least, I thought we had. He had this song. Some band with a name I could never remember. Every few weeks he’d ask if I’d heard it before, and I’d say no, and we’d listen to it again. It was our thing.”

Diane kept her eyes on him.

“That morning he turned it off halfway through. He asked if I ever felt as though people would have an easier time if I wasn’t there.”

He rubbed at a mark on the table with his thumb.

“I told him college was hard. I said everybody had days like that. Then I turned his song back on.”

The thumb moved back and forth.

“I think I pointed at a hawk.”

Diane put her feet down. The floor was empty beneath her. She moved her chair back from the table, crossed to Tom, and knelt beside him.

“I would have listened,” he said. He was looking at the blue paper. “If he’d said it so I could understand, I would have listened.”

“I know.”

“He did, though.”

She put her hand on his arm.

“He did say it.”

In the kitchen, wood struck wood.

Diane turned. The planchette lay at the edge of the counter.

She had left it beside the sink.

It moved another inch. One felt foot reached the edge and went over. The triangle tipped, hung for a moment, and dropped to the floor.

No one had a hand on it. No one was anywhere near it.

The sound it made was small. Tom drew in a breath. Then, from under the counter, came a little wheeze of laughter.

His hand closed on Diane’s wrist.

It came again. The same preliminary hitch, the breath caught before the laugh.

“Ethan?”

Will made a sound in his throat.

Tom let go of Diane and began to stand.

“Don’t,” Will said.

Under the counter the laugh started for a third time. Diane heard it properly now. It began at precisely the same place. Broke at precisely the same place. There was no breath between one and the next.

She got to her feet and switched on the light over the stove.

The planchette lay upside down against a cupboard. Three round pads of felt. A crack in the oak.

The laughing stopped.

Tom was standing so close behind her that she felt his shirt brush her back.

“Do it again,” he said.

Nothing moved.

“Please.”

Claire came past them. She bent with one hand braced on her knee and picked up the wood.

“It’s warm,” she said.

Diane took her elbow.

“Put it down.”

“It’s all right.”

“Claire, put it down.”

Claire held it against her chest. With her other hand she peeled Diane’s fingers from her arm, one at a time.

“You don’t have to stay.”

* * *

Diane stood at the sink while the others sat. Will was saying they needed to end the session. He said it three times, as though repetition might turn the proposal into something they had already done.

Claire put the planchette on the board.

“Then help me say goodbye.”

It sounded reasonable. Marcus sat down. Tom followed, keeping his chair a little away from the table. Will looked at Diane.

She could go back to her room. In the morning she could say she had known better than to stay.

Instead she returned to the fifth chair.

“We’re finishing,” she told Claire.

“Yes.”

Five hands touched the wood.

“Elizabeth,” Will said. “Thank you. We’re going to stop now.”

The triangle stayed still.

“Goodbye.”

It moved toward the bottom of the board. Diane followed the hole as it crossed the linen. Almost there.

It stopped on the first letter of GOODBYE.

Then it went back.

YOU HAVENT ASKED.

“Don’t,” Diane said.

Claire looked straight at her.

“Will I be alone?”

The pointer moved to NO.

Claire closed her eyes. Diane watched the muscles around her mouth relax.

“When it happens. When I die.”

NO.

“Will you be there?”

YES.

“We’ll be there,” Diane said. “I’ll be there. You know that.”

Claire opened her eyes.

“After,” she said.

The kitchen clock went on ticking. Diane had never hated an object as much as she hated that clock.

“I’m frightened,” Claire said. “Every night. I wake up and I think, not tonight, and then I think, what difference does it make which night? I turn on the lamp. I leave the television on. Once I called the pharmacy because they’re open all night and I wanted to hear somebody answer.”

She had said all of this to the board.

“Is she there, Will? Tell me.”

Will’s lower lip moved.

“Something is.”

“That’s enough,” Diane said. “Come away from it.”

The pointer slid to 9. Will pulled his hand away.

“Off. Now.”

Diane let go. Marcus did too. Tom’s fingers dragged over the cloth as he withdrew them.

Claire kept her hand where it was.

“Claire.”

8.

Diane gripped her wrist. She lifted it clear of the board.

The pointer stopped.

For a moment nobody moved. Diane held the thin wrist between her fingers. Claire stared up at her with a naked, astonished hatred.

“Let go.”

“Listen to me. Your mother called him Will. She called herself Elizabeth. You asked her for a song and she didn’t give you one. It said, ‘She told Diane.’ If that was your mother, who is ‘she’? And when Will asked who it was, it gave your name. I think that’s who it wants.”

“Let go of my hand.”

“That was the same laugh. Three times. Exactly the same.”

“You don’t know what it’s like there.”

“Neither do you.”

Claire tried to pull free. Diane held on. Across the table Tom had begun to cry without making a sound.

“You’re hurting her,” he said.

Diane looked down. Her thumb had made a white oval in Claire’s skin.

She loosened her grip. Claire snatched her hand away and put it back on the planchette.

7.

Something knocked beneath the table. All the candles bent toward Claire, five thin flames lying almost flat. Diane caught her own chair before it tipped backward. Its front legs had lifted clear of the floor.

6.

Will went for Claire’s hand. She put her other hand over his and pressed it to the wood.

“Help me,” she said.

He could have lifted her whole arm. Diane saw him measure it: how little she weighed, how easily he could do it. He stayed bent over her.

5.

Tom reached across the table and laid his hand beside theirs.

“If he’s there,” he said.

Nobody answered him.

4.

Marcus caught the edge of the board with both hands and pulled. The frame came toward him an inch. The candle flames stretched longer. Diane could hear something scraping along the floor behind Claire, although the wall was only a foot beyond her chair.

“Pick it up,” Marcus said to Will. “Help me pick it up.”

3.

Diane got hold of the other end. The linen was hot against her knuckles. Together they lifted the board. Claire’s arms rose with it. Her fingers were white from pressing down.

For the first time she looked afraid.

“Mom?”

From directly beneath the board, Elizabeth Maddox said, “Tread softly.”

Marcus let go.

The frame struck the table, and Diane lost her hold.

2.

Claire lifted her hands. Will fell back against his chair. Tom was touching nothing. They all stood or sat with their hands clear, plainly visible in the electric light.

The planchette moved to 1.

The candles went first, all five together, the flames drawn off the wicks like threads. Then the light over the stove. Then the light above the table.

A glass went over and rolled. Marcus said a word twice, the second time from Diane’s side of the table.

Diane still had her hands in front of her. She lowered them until she found the table.

Someone was breathing close to her ear. She leaned away and struck another person’s shoulder.

“It’s me,” Marcus said.

The breathing continued on her other side. It was slow, and there was no warmth in it. It moved the hair above her ear, and it was the temperature of the room.

She could smell the whiskey on Marcus. There was another smell with it: wet earth, and the green, bruised smell of stems cut for a vase. Nobody had brought flowers. She could taste it as well, dry and bitter. She had bitten an acorn once, as a child, to see.

“Phones,” she said.

Nobody answered. A drawer opened in the kitchen. She heard the unmistakable drag of wood on wood, then the small rattle of cutlery.

The drawer shut.

“Who’s over there?”

“Nobody,” Will said.

“Who’s got a phone?”

Claire said, “I’m here.”

Her voice came from the end of the table, where it should have been.

Something below Diane’s hands pressed upward against the tabletop. The pressure traveled along it, slowly, toward Claire. A knee, she thought, and then she thought of how far it had traveled.

Tom began saying his son’s name.

A chair scraped across the floor. It stopped. Scraped again, a shorter distance, as though its occupant was adjusting it to eat.

The kitchen clock resumed. Diane had not heard it stop.

Marcus’s phone lit his face from underneath. Beyond him, Claire sat with her head tilted, listening to someone at her shoulder.

“Hi, Mom,” she said.

The overhead light came back.

There were six chairs around the table.

The new one stood beside Claire. It had a rush seat, blackened with age, and a crooked brass screw through the top rail. Whoever built it had built it out of a single limb, and not a straight one. Diane knew the screw. She had once caught the sleeve of a sweater on it in Elizabeth’s kitchen. She had been wearing the sweater when she sat down and had taken it off, draping it over the chair, because the kitchen was too warm. Coming home from the hospital, she had been cold for weeks. Elizabeth had kept turning up the heat.

Claire put her hand on the rail.

“We threw this out,” she said. “When we sold the house. I watched them load it.”

Tom had turned to look at the floor beside his own chair, and then behind it, as if there might be a seventh.

Will was backing toward the wall.

“Don’t touch it.”

Claire ran her thumb over the screw head.

“Marcus,” she said. “Get me a glass of water, would you?”

He did. Diane watched him go to the sink, fill a glass, bring it to her. When Claire took it, he stood for a moment looking at the empty place in front of the new chair.

“Just one,” she said gently.

He sat down.

* * *

There was no ferry until morning. They kept every light on that night. Marcus tried the cabin’s landline, found a dial tone, and put it down again. He made coffee. Will wrapped the board and carried it outside. Through the window Diane watched him lay the bag in the bed of Tom’s truck, then change his mind and bring it back to the porch. He stood in the rain beside it.

Claire went to bed. Diane followed her and said she would sleep in the room.

“There isn’t another bed.”

“I’ll take the floor.”

“You’ll never get up again.”

Diane fetched blankets from the cupboard. When she came back, Claire was asleep. Her mouth had fallen open. The expression on her face was ordinary exhaustion, and for a moment Diane was able to look at her without fear.

She spread the blankets beside the bed. She lay down without taking off her clothes.

She went back through the evening in order. Five people had sat down to dinner together. By the time the lights went, each of them had been sitting at that table alone.

In the main room, someone moved the sixth chair.

“Leave it,” she called.

The sound stopped.

After a while Marcus came to the bedroom door.

“Did you say something?”

“Stop moving the chair.”

He stood with his hand on the doorframe. Then he looked back into the main room.

“Marcus.”

“I heard you,” he said.

He left the door open.

In the morning Diane sat on the porch with Elizabeth’s letter. Tom had asked her to read it. The paper was drying into little ridges where his hands had held it.

She could not feel the ridges. She turned her hand over. The first two fingers were smooth at the tips, as smooth as the inside of a lip, and when she pressed them to the porch rail she felt the rail in her wrist and nowhere nearer. She had found a thousand lumps with those two fingers.

Near the bottom Elizabeth had crossed out a whole paragraph so thoroughly that the pen had worn through in places. The long tails of her letters stood up out of the scoring. Beneath it she had written: I keep deciding what you can bear to hear. I am doing it even now.

Diane ran her finger beneath the words.

“She never told Ethan she called you,” Tom said from the doorway.

“I don’t know.”

“That’s what it says.”

She looked down at the sheet. “Yes.”

“I keep reading it. I think there’ll be something else the next time.”

Diane folded the page along its weakened crease and handed it back.

“I’m sorry.”

He took it. For a moment she thought he would touch her shoulder. Instead he put the letter in his shirt pocket.

Then he held his right hand out to her, palm up, the way he might have shown her a splinter. The first two fingertips were as blank as hers.

“Marcus has it too,” he said. “He’s keeping his in his pocket.”

Inside, Claire was peeling one of the satsumas. She had set the sixth chair back against the wall. A curl of orange skin hung from her hand.

“Did she always do that?” Tom asked.

“What?”

“Peel it in one piece.”

They watched her remove the last of the skin and lay it on the table in an unbroken spiral.

“I don’t know,” Diane said.

* * *

Claire lived another nine months. Diane drove over from Bend as often as she could, and called on the days she couldn’t. She brought food Claire sometimes ate. She replaced the missing button with a little blue one because the wooden fish were no longer made.

Marcus returned the emerald. Tom answered his messages. Will said he was going to burn the board, and when Diane asked what he was waiting for, he said a dry day.

There were good afternoons. Once, before Christmas, all five of them sat in Claire’s living room and complained about the neighbors while Claire dozed through the best parts. One neighbor had a bird feeder shaped like a lighthouse, and Tom was offended by it. He called it a tacky garden ornament. Marcus argued the other side for sport, and the longer he went on, the less he believed it, until Claire opened one eye and asked whether he owned one. Diane held on to that afternoon. She remembered the crumbs on the sofa, the volume of the television, the way Claire reached up without looking and Tom put a biscuit into her hand.

Will got his dry day in January. He burned the board and the pointer in a steel drum behind his house in Olympia, and a neighbor two doors down said afterward that the smoke had smelled like a florist’s.

Tom was the one who telephoned. Diane knew before he spoke, because Tom did not telephone.

“They found him this morning,” he said. “In his kitchen. They’re saying his heart, and that the house was very cold.”

“Was he sitting down?”

“Diane.”

“Was he?”

“Yes. At the table.” She heard him breathe. “There were three chairs in that kitchen. His sister says he only ever had two.”

“What did the third one look like?”

“I didn’t see it. She said it was old. She said the seat was made of something like straw.”

That night Diane called Will’s number anyway and let it ring. Nothing in his house answered.

The chair came up in March.

Claire was on the telephone. Diane was washing a bucket in the barn.

“I’m having it brought over,” Claire said.

Diane turned off the hose.

“From Will’s.”

“His sister was going to put it out with the trash.”

Diane had last seen that chair against the wall of the cabin in November. Nobody had driven it to Olympia.

“Claire.”

“It’s a chair. I can have my mother’s chair.”

The bucket knocked against Diane’s boot. She gripped the phone so hard that its edge hurt her ear.

“It was in his kitchen when he died.”

“So was the kettle.”

“I don’t think you should.”

Claire sighed. “You never have to sit in it.”

By April it stood at the head of Claire’s kitchen table. Diane could have carried it out to the car on any afternoon she liked. She drank her coffee at the counter.

In June she let herself in and heard singing. Claire was at the sink with a towel over one shoulder, making a noise that rose without warning and came down somewhere else.

“You never liked that,” Diane said.

“Didn’t I?” Claire turned around. She looked rested. “Sit down, sweetheart. You’re hovering.”

Diane stayed on her feet.

Claire was sleeping through the night by then. She said so herself, and seemed to think it was good news.

At the hospital in August, the chair was beside the bed the first time Diane went in. Rush seat. The crooked brass screw. Nobody at the nurses’ station could say who had signed for it, and one of them asked whether it was a religious thing.

She did not sit in it.

Claire looked at her for a long time.

“What?” Diane asked.

“Nothing.”

She sat on the edge of the bed.

On the last afternoon Claire was awake, Diane read to her. A paperback of Yeats lay open across her knees. She had marked the poems Elizabeth used to read, but Claire wanted a different one, then another, then asked her to stop.

“Sorry. I can’t follow it.”

“We don’t have to do anything.”

Diane closed the book. Down the corridor someone laughed, and Claire turned her head toward the door.

“Diane.”

“I’m here.”

“I told you that you didn’t have to stay.”

“At the cabin?”

Claire nodded.

“You stayed.”

“Yes.”

Her hand came across the blanket. Diane took it.

“How do you know when it’s somebody?” Claire whispered.

Diane leaned closer.

“What do you mean?”

Claire was looking past her at the chair.

“Sometimes I forget and say things. Things I was going to tell you. Then you come and you don’t know them.”

“Say them to who?”

“You. Last night. You sat in that chair and held my hand.”

“I was in Bend last night.”

“I know that now.”

Diane’s hand tightened around hers.

“What things?”

“It doesn’t matter.” Claire frowned. “You told me things too. There’s a little boy. He’s been there the longest. He’s only got the one boot.”

“Claire.”

“He counts when anybody comes for you. He gets to the end, and then he’s sorry. There’s a woman who won’t get up out of her chair. Mom’s there. Will.”

Claire looked at the chair again. With her free hand, Diane reached out and pushed it back against the wall.

“Look at me, Claire. I’m here now.”

“I know.”

“I’ll stay.”

Claire held her hand. Gradually the grip loosened, then tightened with startling force.

“Don’t let her take me,” she said.

Diane stood so quickly the book fell from her lap.

“Who?”

Claire’s eyes were open. Her face had emptied of everything Diane knew how to answer.

“I don’t want to go. I said I did, but I don’t. Tell her.”

Diane pressed the call button and kept her thumb on it.

“It keeps telling me I won’t fall,” Claire said.

The nurse came, asked Claire where she hurt, checked the pump, stayed until her breathing settled. Diane stood on the other side of the bed with Claire’s fingers locked around two of hers. They were the two she could not feel. She could not seem to persuade anyone that the words mattered.

Claire fell asleep. Diane stayed until she woke.

“How’s Phoebe?” Claire asked.

Diane began to cry.

“Oh, sweetheart.” Claire reached up and touched her cheek. “It’s all right now.”

Behind Diane, against the wall, the rush seat creaked and settled.

Claire looked over Diane’s shoulder.

“Thank you,” she said.

* * *

After the funeral the three of them went back to Claire’s house, because there was food to finish. Somebody had brought the chair home from the hospital and stood it at the head of the kitchen table. Nobody admitted to that either.

Tom found Diane at the sink. He took the letter out of his shirt pocket and laid it on the counter.

“You can read her writing,” he said. “I’ve had it under a lamp since January. I kept thinking I’d get it. I can’t.”

The paragraph was scored through, the pen gone into the paper in places. But the tails of Elizabeth’s letters came up above the scoring and hung down below it, and Diane had been reading that hand upside down across a breakfast table since she was twenty. She went slowly, a word and then the next word. Tom wrote them on the back of the order of service.

“‘In thirty years I never once let any of you sit down by yourself in that house,’” she read. “‘You thought I was being hospitable. There is something in that garden that keeps company with whoever sits alone, and it does not give them back.’”

Tom’s pencil had stopped.

“Go on.”

“‘The chair by the window came with the place. I did not care to see it empty. The woman before me had it made from the limb her boy’s swing hung on, after the limb came down with him on it. She had called from that window that he might have ten more swings, and he was counting them. It still counts, if you come to take someone away from it. At the end of the count it holds on.’”

“There’s one more line,” Diane said.

“‘I left your son sitting under that tree for an afternoon, and I went in to see to the lunch.’”

Beneath that, uncrossed, was the sentence they had both read a hundred times. I keep deciding what you can bear to hear. I am doing it even now.

Diane looked at the chair at the head of Claire’s table. She thought of April, and May, and the counter where she had stood with her coffee. In all that time she had not once sat down in it. Elizabeth would have.

They burned it that evening behind Claire’s garage, Tom and Diane, with Marcus watching from the back step and his right hand in his pocket.

“Will burned the board,” Marcus said.

“Will was by himself,” Diane said.

It burned the way a chair burns. The smoke smelled of wet earth and cut stems, and none of them said so.

“If he’s in there,” Tom said. He was watching the rush seat go. “If I’d known he might be in there, I’d have sat in it.”

Diane heard that. She said, “I know,” and raked the fire together, and let it pass.

* * *

When she got home, Diane told Phoebe everything.

She told her at the kitchen table, in daylight, with both their phones lying between them. She showed her the two fingers. Phoebe listened. When Diane got to the hospital, Phoebe took her hand and waited until she could go on.

“I should have told you before,” Diane said.

“You’re telling me now.”

They put a lamp on the landing. Phoebe left the bedroom door open because Diane wanted to hear the ordinary sounds of the house. At night, when she woke, she would listen for water running through the pipes or the low fuss of the alpacas beyond the barn.

For three weeks this helped.

Then, a little after two one morning, a chair moved in the kitchen.

Diane opened her eyes. The landing light lay in a narrow strip across the bedroom floor. Phoebe slept beside her, one hand caught under her pillow. The room smelled of wet earth and cut stems, and her mouth tasted of acorn. They had burned the chair. She had raked the ashes herself.

The chair moved again.

“Diane?”

Claire’s voice. Tired, a little impatient. The voice she used when she had been waiting on the phone while Diane finished something that was supposed to take a second.

Diane sat up.

“I know you’re awake.”

She put her feet on the floor. The boards were cold. She remembered the fifth chair in the cabin, the touch against her leg. Phoebe was asleep and Diane was not, and that was all it had needed.

In the kitchen, Claire coughed.

Diane reached for Phoebe’s shoulder and shook it.

“Wake up.”

Phoebe stirred.

“Wake up. Don’t go downstairs.”

“What is it?”

Diane held on to her. She heard herself begin to say Claire’s name and stopped.

“There’s something in the kitchen.”

Phoebe sat up. They listened together.

“I can hear you,” Claire said. “I came all this way.”

Phoebe reached for her phone.

The chair scraped a short distance. It was one of their own kitchen chairs. The old one was ash, and burning it had changed nothing. It had never needed that chair. It would sit in any empty chair, next to anyone who was alone. Diane could picture the chair being pulled closer to the table. She could picture a hand laid flat, the little pause before Elizabeth required everyone to be kind.

“No,” she said.

The house was quiet.

Then Diane’s voice came up from the kitchen.

“Phoebe? Come here a minute.”

Phoebe stopped with the phone in her hand.

The voice was very good. It had the slight strain Diane always heard in recordings of herself, a complaint she never heard while she was speaking.

“I need some help.”

Diane pulled Phoebe against her. She felt the phone trapped between them, the hard corners against her ribs.

Downstairs, in Diane’s voice, something began to cry.

They stayed on the bed. Diane kept one hand on the back of Phoebe’s head. With the other she reached for the lamp and turned it on.

The crying stopped.

The chair scraped once more across the kitchen floor, and stopped.

They did not sleep. They sat against the headboard with the lamp on and their shoulders touching, and whenever one of them began to nod, the other said her name. Nothing came up the stairs.

When it was light they went down together. One of the kitchen chairs stood pulled out from the table and turned to face the hall. The smell was gone.

“It couldn’t,” Phoebe said. “Not with both of us.”

“No.”

“So where does it go?”

Diane was already looking for her phone.

Tom answered on the second ring.

“You’re up early,” he said.

He sounded rested. She had not heard him sound rested in eight years.

“Are you at your table?” Diane asked.

“I’ve got someone here. I can’t really talk.”

Behind his voice, someone laughed. There was the small hitch, and then the wheeze. Then it came again, from the hitch.

“Tom. How many chairs are at that table?”

“Two,” he said, and he sounded proud of it. “I got the other one down from the garage.”

“Stand up. Go out in the yard and stay on the phone with me. I’m coming to get you.”

“You don’t need to do that.”

Phoebe already had the keys.

It was three hours to Tom’s. Phoebe drove. Diane kept him talking as far as the pass, where the signal went, and after that she held the phone against her ear anyway.

His truck was in the yard. The blinds were down in every window. On the porch Diane smelled it before she touched the door: wet earth, and the green of cut stems, stronger than she had ever known it.

“Stay in the doorway where I can see you,” she told Phoebe. “If I sit down in there and can’t get up, come and get me. And if you hear me tell you to leave and my mouth isn’t moving, I didn’t say it.”

“That’s quite a sentence,” Phoebe said. She stayed in the doorway.

Tom sat at the kitchen table, facing the window. His right hand lay open on the wood, palm up, the fingers a little curled, and he was looking into it. Drawn up beside him, so close that the frames touched, was the chair he had brought in from the garage, a folding chair with a green canvas seat. No one was in it, and the canvas hung down in the middle as if someone was sitting back in it.

“You made good time,” Tom said. “Ethan, it’s Diane.”

He waited. He nodded at something.

“He says hello.”

Diane had heard nothing. She could taste the room now, dry and bitter at the back of her tongue. The acorn again.

“I need to sit down, Tom.”

“There’s only the two.”

“I know.”

She had stood at Claire’s counter long enough. She turned and sat in the folding chair.

The canvas was cold through her jeans, the way the chair from the bedroom had been cold on Orcas. It took her weight and went on taking it. What was under her did not move over. Something breathed out against the side of her neck, and there was no warmth in it.

Close to her collarbone a child said, “Nine.”

It was a boy’s voice, clear and a little bored.

“He hasn’t done that before,” Tom said. “I expect that’s you. I know what it is. I was standing next to you when you read it.”

“Then you know what happens at one.”

“I do.” He had not looked up from his hand. “I think I’ll stay for it, Diane. If it’s all the same to you.”

“Eight.”

“Ask him what the band was called,” she said.

“What?”

“The song in your truck. You never could remember the name. He must have told you fifty times. Ask him.”

“Don’t do that.”

“Ask him, Tom.”

“Seven.”

The canvas had drawn tight across the backs of her legs.

Tom raised his eyes. He looked at her, and then she understood that he was looking at the chair and that she happened to be in it.

“Eth,” he said. “What was that band? The one I never could get.”

The laugh came from just behind her ear. The hitch, and the wheeze.

“Come on,” Tom said. He was smiling. “You told me every time.”

“Six.”

Then it was Ethan’s voice.

“You never liked my singing,” it said.

Tom’s hand closed on the table.

“It told Claire that,” he said. “On the island. A letter at a time.”

“Five.”

The cold had reached her ribs.

“Look at me,” she said.

He did.

“He turned off his own song in the truck to ask you whether people would have an easier time without him. You were there for that. Whatever this is, it wasn’t.”

“Four.”

“We never finished saying goodbye. It wouldn’t let us. Say it now.”

“I don’t want to.”

“I know.”

“Three.”

Tom looked at the window blind, at the slats and the daylight behind them.

“Goodbye, Ethan,” he said to the window.

Diane took him by the wrist and stood. For a moment the chair held her. She pulled, and it pulled. Then something gave, like a sleeve tearing, and she was on her feet and he was on his. She knew that she was hurting him. She did not loosen her grip.

“Two.”

The boy’s voice was behind her now. It came from the empty chair.

“Phoebe,” said Diane’s voice from the corner by the stove. “Go and wait in the car.”

Phoebe looked at Diane’s mouth. She held the door wide and did not move.

Diane made for the door and dragged Tom with her. It was six steps. He came like a man being walked out of a deep sleep, heavy, his boots catching on each other, and his chair went over behind them. Her shoulder hit the doorframe. Phoebe got a fistful of his shirt and hauled.

“One.”

They came through the door in a heap and went down on the boards of the porch, all three of them, and that was when it came. Behind them the kitchen made the sound a big limb makes when it lets go of the tree. It started with a tick. Then came a long green tearing, and then the whole weight of it crashed down onto the kitchen floor. The porch jumped under them. The blinds slapped against the windows. Somewhere in a cupboard, plates broke. Phoebe had both arms over her head.

Then it was quiet. When Diane looked back through the door, the table had jumped against the wall, and the folding chair lay flat under nothing, bent the wrong way at every joint. At the count of three she had still been sitting in it. The limb had come down at the end of the count, the way it came down with the boy on it, and this time all it got was a chair.

Tom tried to get up and could not. He got as far as the top step and sat down on it. Phoebe sat on one side of him and Diane on the other. Nobody said anything. Phoebe was shaking, and after a minute Diane was too.

After a while Diane saw that she still had him by the wrist. She could see her hand around it. She could not feel his wrist, or anything else, with that hand. Until that morning it had only been the two fingertips.

In the kitchen a woman none of them had ever heard said, “Ten more, Billy. Then you come in.”

* * *

Tom took the spare room. He brought two bags and the letter and nothing else out of that house, and at Christmas he was still in the spare room. Marcus came on Fridays and left on Mondays and was, Phoebe said, the only guest she had ever known to bring his own towels. In Tacoma he let his own spare room to a law student, at a rent that puzzled her.

Diane told them all of it, at the kitchen table, in daylight, and read them Elizabeth’s paragraph from the back of the order of service. She did not decide for anyone what they could bear to hear.

* * *

At Diane Weiss’s house, no one was allowed to sit alone. A chair would be found, a place made, a little more water added to the soup; Diane could accommodate anyone, and for thirty years nobody who ate there mistook this for a house rule. There was a man who came for one night and remained for nineteen years. There was another who came on Fridays until he could no longer drive. Outside the kitchen window stood an oak that no one had planted. It had grown around a strand of pasture wire and the head of a rake that Phoebe had leaned there and forgotten, and Phoebe mowed around it for as long as she could mow.

They went from that house one at a time, with the chairs full. Phoebe was the last of them but one.

On an evening in late summer Diane sat down at the kitchen table by herself, because there was no one left to sit down with. The long shadow came in through the window. A chair was found.

What it could not move, it made room for.

Share

Subscribe now

My mother has seen my handwriting. Against all available evidence, she still believes it belongs in her greeting cards – preferably in cursive.

It’s not that I don’t appreciate a handwritten card. Quite the opposite. I understand why Mom loves them. There’s something intimate about recognizing someone’s handwriting, knowing they sat down with a pen and took a moment to write something just for you. To her, those little marks on the paper are evidence that a real person cared enough to make the effort.

The problem is that my handwriting has spent so many years in retirement that even I have trouble reading it. So I make my cards digitally. And I don’t mean picking a generic template, typing “Happy Birthday,” and calling it a day. I mean designing the card, choosing the artwork, incorporating personal photographs, getting the layout just right, and spending an unreasonable amount of time rewriting a message until it actually says what I want it to say. I can even make the typography look handwritten, although apparently that’s where the fraud begins.

Mom, meanwhile, can walk into Target, pick out a card designed by someone she’s never met, containing a sentimental message written by another complete stranger, add a few handwritten sentences, and consider the result personal. I can spend an entire evening creating something specifically for her, filled with our memories and words I’ve agonized over, and she’s still a little disappointed because the words came out of a printer.

I love my mother. I even appreciate the tradition she’s trying to preserve. But there’s something wonderfully upside-down about the idea that a mass-produced greeting card becomes personal when you add a few pen strokes, while something conceived, designed, and written entirely for one person somehow loses its personal credentials because a computer was involved.

To me, the thought is the personal part. The medium is simply how I deliver it.

And lately, I’ve been noticing a remarkably similar disagreement in the world of software engineering.

Many of my friends and colleagues have spent most or all of their careers as software engineers. Like nearly everyone else in technology, they’re figuring out what it means to work with AI now that the question has moved beyond whether to use it and toward how completely it should become part of the way we work. We use many of the same tools, talk enthusiastically about many of the same possibilities, and occasionally use the same phrase – AI-native engineering – to describe what we’re doing. The more we talk, though, the more I suspect we’re using the same words to describe two fundamentally different journeys. We’re heading toward roughly the same place, but we’re approaching it from opposite sides of the fence.

For many traditional software engineers, the journey begins with implementation. These are people who spent years learning programming languages, algorithms, frameworks, runtimes, databases, types, memory, networking and all the wonderfully specific things computers have historically required humans to understand before they’ll reliably do what we want. AI enters that world as an extraordinary accelerator. The engineer who already knows how to write the code can now describe some of it instead. The engineer who once spent an afternoon implementing a routine can ask an agent to produce a first version in minutes, inspect it, correct it and move on. The interface to the work changes dramatically, but the engineer’s underlying model of the work doesn’t necessarily change with it. AI allows them to spend less time manually implementing things they already understand how to build.

I seem to have arrived from the other direction. I started my career as a software engineer, but the parts of the job that interested me most were always closest to the human being on the other side of the screen. I loved figuring out what a product should do, how information should be organized, what an interaction should feel like, how one system should connect to another, and how something technically complicated could become simple and intuitive when it reached the person actually using it. I could write code, and I did, but I never fell in love with the ritual of coding itself. I didn’t enjoy staring at thousands of lines of implementation, mentally tracing my way through loops and branches, or spending an afternoon discovering which obscure technical detail was responsible for something behaving differently than I intended. Eventually my career moved toward product, experience, architecture, creative technology, media, and leadership – not because I stopped enjoying building things, but because I moved closer to the parts of building that I actually loved.

That instinct predates my career in big technology companies. Years ago, when I was building websites for small businesses, the work began long before I opened an editor. Many owners were still deciding whether the Internet mattered to their business at all. I would sit with them and talk about what the company was, who its customers were, what those customers needed, what the website should communicate, and why anyone would want to use it. Then I would design the experience, create the visuals, and build the site. Front end, back end, design, customer conversation, and business purpose were not separate departments in my head. They were all parts of the same thing. And I loved that.

When I later joined a much larger engineering organization, I discovered how thoroughly mature software companies can divide that whole into specialties. I remember arriving expecting some version of the work I had already been doing – design the experience and then build it – only to discover design belonged to another team. My job was implementation. The division made organizational sense, but something important had disappeared for me. I was no longer making the whole thing; I was making my assigned portion of it. Looking back, I suspect that was at least as important to my eventual burnout as the thousands of lines of code themselves.

For most of software history, there was a practical reason implementation expertise held so much power. You could understand a customer beautifully. You could design an extraordinary experience, map every workflow, anticipate every edge case, and know precisely what the product should do. Eventually, someone still had to write the code. Implementation knowledge wasn’t merely one useful skill among many; it was admission to the factory. If you couldn’t translate the idea into the languages and frameworks computers understood, you either needed someone who could or you weren’t shipping much of anything. That gave implementation expertise an unusual authority over the entire act of software creation, even though knowing how to implement a product has never been quite the same thing as knowing what product ought to be implemented.

I learned that distinction early. As a full-stack engineer with a focus on the front-end, I occasionally encountered pages and interfaces that had been created primarily by backend engineers. Sure, they worked. Every necessary piece of information was technically present. Buttons could be clicked. Data went where it was supposed to go. They were also, to put this charitably, not particularly well designed from a visual or usability standpoint. The people who built them could run circles around me in areas of computer science that simply weren’t where my own interests and strengths were, but none of that knowledge automatically told them where a button belonged, which information deserved visual priority, why an interaction felt awkward, or why a technically functional page could still be frustrating to use. Implementation expertise and product judgment overlapped, but they were never the same thing.

AI hasn’t created that distinction. What it has done is make it possible to cross it in ways that weren’t previously practical. For the past year and a half, I’ve been building complete software products again, but I haven’t returned to the way I built software twenty years ago. I work conversationally with AI agents. I define what the product needs to do, establish architecture and constraints, describe the behavior I expect, interrogate decisions, test results, reject approaches I don’t like, and keep pushing until the thing behaves the way I believe it should. If a field should accept only a particular kind of information, that’s the contract I care about defining precisely. If a duration needs one level of precision internally while displaying another to the user, I care deeply about specifying that behavior correctly. If an operation must continue after the user leaves a page, if two records must change together or not at all, or if a user must never be able to manipulate an identifier and see somebody else’s data, those are requirements I am exceedingly particular about. What I increasingly don’t care about is personally remembering every language-level mechanism used underneath them.

That can sound cavalier, but only if we assume rigor must reside in the same place it always has. Mine hasn’t disappeared; the mechanism producing it has changed. I can ask one agent to implement a specification and another to review the implementation without touching it. I can ask a third to search specifically for failure cases, another to examine security boundaries, and another to generate tests designed to break the assumptions the first agent made. I can isolate agents so they don’t inherit one another’s conclusions, ask for competing implementations, force each approach to explain its tradeoffs, and build independent checks for results that matter. The human responsibility shifts from personally performing every implementation and review task to designing a process in which implementation, criticism, testing, and verification can challenge one another. “Trust me, it compiled” was never a sufficient quality strategy for human-written code; “trust me, the model said it works” isn’t one either.

This is where I think the phrase AI-native begins to split into two meanings. For the traditional engineer, AI often performs work the engineer already knows how to do. If a requirement changes, that engineer may already know exactly how the algorithm needs to change and could write the modification by hand; AI simply makes the change faster. That’s powerful, but it is primarily an acceleration model. The engineer still knows how to do the work AI is doing.

For the AI-native builder, that isn’t necessarily the bargain. I may know precisely what I need the software to do without knowing precisely how I would implement it. If I decide that a process now needs to track how many times a state changes, I don’t necessarily need to stare at the existing function until I can personally devise the new variables and control flow. I can specify the new behavior, ask an implementation agent to modify the system, require an explanation of the change, add tests for zero, one, and multiple transitions, have another agent review the implementation, and verify that the new capability hasn’t changed the original result. AI isn’t merely accelerating an implementation I already know how to perform. It is extending my ability to build beyond the boundary of what I personally know how to implement.

That distinction matters because a rule that says you may delegate work to AI only when you could have done the work yourself would neuter one of the most consequential properties of the technology. We already accept AI extending people into disciplines they haven’t mastered. Software engineers now use AI to generate interface concepts, copy, graphics, and front-end experiences without first becoming professional visual designers, writers, or creative directors. The results vary wildly, because generating something and exercising good judgment about it are different skills. But the premise itself rarely seems shocking: AI can extend an engineer outward into adjacent disciplines. It is curious that the reverse movement – a product, design, or architecture person using AI to move inward into implementation – is still so easily treated as suspect unless that person can prove they could have written the implementation manually.

This creates an interesting contrast in what each group considers precision. Imagine that we’re both building a feature involving a numerical value. A traditional engineer’s questions may naturally travel downward: What type is this? How is the decimal represented? What happens if an integer becomes a floating-point value? How does the runtime perform this comparison? My questions tend to travel outward: Why is the customer entering this number? What precision actually matters to the person using it? What should we store, what should we display, and should those be the same thing? What happens at the boundaries? Where else does this value appear in the product? What would make the result misleading or confusing? Both are forms of precision. They’re simply aimed in different directions. The traditional engineer’s precision often travels deeper into implementation. Mine tends to travel further into the product.

AI has not made me want to become the kind of software engineer I deliberately stopped being years ago. What it has done is remove an activity I actively disliked from the critical path between I can conceive this product and this product exists. That lets me spend more of my finite attention on architecture, product behavior, user experience, visual design, workflows, feature strategy, integration, language, failure behavior, and the thousand small decisions that determine whether something merely functions or actually feels like a coherent product. I don’t want someone to hand me a Python repository and say, “Here’s your piece.” I also don’t particularly want to design an experience and throw it over a wall to an engineering team. I want the thing – the whole product – and AI increasingly lets one person retain ownership across boundaries that organizations historically had to divide among specialists.

That doesn’t mean those specialties disappear. It means one person can increasingly direct capability across more of them. I can move from customer problem to product strategy to interaction design to architecture to implementation direction to testing to deployment to positioning without pretending that I’ve suddenly become the world’s deepest specialist in every one of those areas. The connective tissue is no longer the programming language. It’s the product. The human increasingly holds the intent, judgment, and coherence while specialized machine collaborators supply capabilities on demand.

This is why I find some of the current conversation around “vibe coding” simultaneously useful and inadequate. There absolutely are people asking an AI to make something, accepting the first output that appears, seeing that it runs, and calling the job finished. That’s not what I’m describing. The difference lies in what happens between the first AI-generated result and the moment someone decides the work is ready to ship. Traditional engineering already relies on layers no individual engineer personally recreates – compilers, databases, operating systems, frameworks, cloud platforms, automated tests, static analysis, and countless dependencies. We learned long ago that correctness should not depend entirely on what one programmer happens to know and notice. AI extends that abstraction into implementation itself.

The industry hasn’t quite figured out what to make of that yet. We still tend to define technical credibility using the competencies of the people who historically had exclusive admission to the factory. Even supposedly AI-native conversations can quickly return to syntax, language internals, framework minutiae, and implementation details that made perfect sense when the human was the implementation engine. Some of that knowledge will remain valuable, and in some domains indispensable. But the more consequential question is no longer whether every builder can personally reproduce every implementation. It is whether the system they direct can reliably produce the intended product, expose its assumptions, survive meaningful testing, and provide enough evidence to deserve trust.

At the same time, implementation is becoming dramatically cheaper, and that changes the relative value of everything surrounding it. When code becomes cheaper, deciding what deserves to become code becomes more valuable. When technically functional interfaces can be generated in minutes, visual judgment matters more. When features become inexpensive to implement, product judgment matters more. When ten plausible solutions can be produced before lunch, knowing which one actually serves the customer matters more. AI can generate an interface astonishingly quickly; it is still far less reliable at looking at that interface and recognizing that the hierarchy feels wrong, the interaction is irritating, the language doesn’t sound human, the page feels cheap, or the requirement was satisfied while the person who has to use it was forgotten. Those are not secondary concerns to me. They are the product.

Perhaps this means we’re trying too hard to squeeze the emerging role into the word engineer at all. “AI-native product engineer” gets closer. “AI-native builder” may be closer still. I’m less interested than I used to be in defending the title. What interests me is the expanding class of people who can now move from conception to functioning software without first becoming specialists in every implementation layer between those two points.

For decades, implementation knowledge was admission to the software factory. AI is opening its doors to people who have always known what they wanted to build but didn’t necessarily possess every specialized skill required to build it themselves. I still read code, challenge technical decisions, test behavior, demand explanations, and take responsibility for what comes out of that factory. I simply don’t believe personally writing every line should remain the price of admission. What excites me is the ability to conceive something meaningful, shape the entire experience, and stay involved until the finished product becomes real.

As for Mom, I suspect we’ll continue our greeting-card disagreement for years to come. She’ll keep appreciating the warmth of a handwritten message, and I’ll keep designing cards filled with photographs, personal memories, and carefully chosen words. I rather like that neither of us is likely to change our mind.

But if you’re a software engineer who wants someone to appreciate every line of code you wrote by hand, I know exactly who would love to receive your Christmas card.

Just make sure you write it in cursive.

Share

Subscribe now

Every summer delivers two things on schedule: a soul-sucking heat wave, and a monster. The most famous summer monster ever built was a mechanical shark named Bruce, after Spielberg’s lawyer – no relation, on either count – and the beautiful part is that Bruce barely worked. Salt water ate his hydraulics. He sank to the bottom of the sound on his first day in the ocean. He broke down so relentlessly that Spielberg had to shoot most of the movie without him, which, ironically, is exactly why the movie terrified the planet: two notes on a tuba and the hint of a fin emptied every beach in America. The monster almost never functioned, but the fear never broke down once.

This summer got its monster too, and it’s faithful to the original in the one way that matters: it barely needs to exist. It’s two letters long, and it bites you in the analytics.

Sometime in the last few weeks, LinkedIn started writing people up – quiet notes informing you that readers reported your post because it seemed like AI slop.

Not that it was AI slop. Nobody has to establish that. Nobody has to establish anything. The report button shipped at the end of July, hiding under the three-dot menu like it had always been there, and it does not say “This is AI-generated.” It says the post seems like AI slop. A report button for a feeling. Read that word again, because the entire machine balances on it. “Is” would be a claim, and a claim can be wrong. “Seems” cannot be wrong. You can’t disprove somebody’s impression of your writing – and this summer, their impression got machinery. A million members clicked it in the first three weeks, every one of them quietly feeding the software that decides what the rest of us see next.

There’s a scene in that movie everybody remembers. Fourth of July, beaches open, the mayor personally coaxing a family into the surf so the tourists will follow. Brody has already sent his own son to the pond, a sleepy backwater the boy complains is for old ladies. Do it for the old man, his father says. Then a fin cuts through the swimmers, and a packed beach empties in about ninety seconds. Amity answers with everything it has. Patrol boats converge. Deputies lean over the side with rifles trained on the fin, and they are about to fire when two terrified boys pop up holding a piece of cardboard. Guns drawn, on cardboard, held by children. And while every eye and every barrel on that beach points at the fake fin, the real shark slips through the inlet into the pond – the water for old ladies, the water the police chief himself chose because it was safe – and takes a man off a rowboat.

That scene – both halves of it – is this entire summer. The panic requires no verification. The enforcement, when it shows up, shows up armed and aims at the wrong thing. And the real bite happens in the boring water nobody is watching.

Because here’s the thing about monster movies: the monster is never the story. The town is the story. Jaws is not about a shark – it’s about Amity, a town deciding what it’s willing to see. So this isn’t really an essay about AI.

It’s about the town.

* * *

There’s another scene people forget. Halfway through the movie, the town catches a shark. They string it up on the dock and pose for pictures, and the mayor declares the beaches safe. The scientist wants to cut the shark open to be sure it’s the right one; the mayor won’t hear of it – not with cameras on the dock. Verification is bad for tourism. It is, of course, the wrong shark. The one doing the biting is still in the water.

That’s the whole argument, right there on the dock. Slop is real – the feeds genuinely are filling with careless, worthless posts, and I’m not here to tell you your eyes are lying. But the town has strung up the wrong animal. Slop was never an AI invention. Bad writing predates the chatbot by about five thousand years. So does spam. So does plagiarism, padding, phoning it in, and publishing garbage while calling it done. AI made carelessness faster and nearly free, and the flood that followed is real – but a flood of an old thing is not a new thing. The monster the town wants is the machine. The monster it has is the one it’s always had: a person who hit publish without caring what went out.

AI doesn’t publish slop. Humans do.

Everything else in this piece – the buttons, the scanners, the invisible marks, the write-ups – is the town on the dock, posing with the wrong fish.

* * *

Let’s meet the town council. The mayor of Amity spends the whole movie in a blazer printed with little anchors, reminding reporters that Amity means friendship, keeping the beaches open because summer dollars are summer dollars. LinkedIn’s entire brand is connection. Friendly towns with something in the water reach for the same playbook: deny the problem while the money is swimming, then, once the crowd turns, punish it as theatrically as possible.

So here’s LinkedIn’s product chief, this summer, mid-turn. He says AI slop is “a top priority for all of us.” He also says “AI and slop are not the same thing.” He also says “slop is hard to define and the definition changes” – adding, helpfully, that your reports let them tune their models. Stack those up and enjoy the view. It’s a top priority. It isn’t the same thing as AI. And nobody can define it – but the reports become the definition, and the enforcement is already live. Definition later. Harpoons now.

Oh, and in the same announcement, LinkedIn killed “Enhance your post” – the AI writing assistant LinkedIn itself bolted into the compose window, the one that spent two years helping fill the very feed now under suspicion. The company didn’t just step on the rake. It manufactured the rake, handed one out free with every profile, and then installed a button for reporting anyone seen holding one.

One more detail from the fine print. LinkedIn says it built safeguards so that individual feedback can’t be used to “unfairly target other members.” Sit with that. They poured the guardrail before most members knew the button existed. You don’t pour concrete at the edge of a cliff nobody is driving toward.

* * *

When Amity finally admitted it had a problem, it didn’t send for experts. It posted a bounty – and by morning the harbor was wall-to-wall amateurs in overloaded boats, jostling for a shot at a monster none of them could identify. That’s the button. A billion members deputized overnight, no training required, no definition supplied.

And since nobody agrees on what slop is, everybody hunts by folklore instead: em dashes, emoji bullets, thirty one-sentence paragraphs in a row, the word “delve.” The surface tics get reported, never the substance. It is the smallest hill in publishing, and people are dying on it daily. The reports train the detection software, and the verdict now costs real money: LinkedIn says members are seeing 40% less of what it classifies as AI slop than they did a few weeks ago – a number the company brags about while declining to say how much your clicks actually caused. So writers do the only rational thing – they scrub the tics. And here the loop closes on something genuinely strange. “Human writing” quietly stops being something the software detects and becomes something the software defines. Whatever doesn’t get flagged, that’s human. A property invented by the button, with a price tag attached.

I’m Exhibit A. I banned em dashes from my own writing months ago. Not because anything is wrong with an em dash – I used them happily for decades – but because I now publish into a world where a stray dash is treated as evidence, and I’d rather lose a punctuation mark than argue with a feeling. You won’t find a single one in this essay. I wish I could tell you that was a style choice.

* * *

Here’s where I’m supposed to concede that all of this is really about quality. So let’s test that.

Long before AI entered the picture, I quietly accumulated more than a dozen free years of credit monitoring – parting gifts from companies whose human-coded, human-designed, human-approved apps leaked my personal information to hackers. My identity has been stolen twice. Both times, the ceremony was identical: an email mentioning that my records had wandered off along with a few hundred million other people’s, a coupon for credit protection, and a warm suggestion that I contact the bureaus. That was it. That was the whole reckoning. No mark. No scanner. No reach penalty. There is no “seems like negligent engineering” button on the App Store.

An em dash on LinkedIn now draws more organized outrage than a database breach. The breach got a news cycle and a settlement; the punctuation got machinery. And I understand why, sort of. You can spot an em dash from your towel in two seconds. You cannot see careless engineering from anywhere on the beach. Outrage flows toward whatever’s easy to see, not whatever does the damage. Sharks got a theme song. Riptides just get you.

Whatever safeguards those companies had, my information got out. Humans built the apps, humans shipped them, and I got the credit-monitoring coupon – years before there was a chatbot to blame. The breaches were the pond. The real bite always happens in the water everyone agreed was safe, while the boats surround a fin.

* * *

So here is my definition of slop, and notice that it never mentions the tool: slop is what happens when nobody sets a standard going in, nobody checks the work while it’s being made, and nobody reads it before hitting publish. Made without care, reviewed by no one, shipped anyway. The emoji-bullet spam everybody pictures absolutely qualifies – not because an AI touched it, but because no human did.

Run two essays through that test. One is brilliant, argued to the bone, written with an AI in the room the entire time, draft after draft. The other is terrible, hand-typed, every cliché lovingly selected by human fingers. The “AI touched it” standard convicts the first. Mine convicts the second. The internet convicts whichever one it’s mad at today, and calls that a standard.

I recently sat a timed technical interview built entirely around working with AI – using the tools was the point. In the first few minutes I had the machines generate quick, working answers, which I saved as insurance in case the clock or the platform betrayed me. Then I spent the next five hours doing the actual work: running multiple AIs against each other, making them critique each other’s output, rewriting, testing, throwing drafts away, until the solutions were something I’d sign my name to. The hours weren’t what made the work better. The decisions I made during them were. Same tools produced the panic drafts and the finished work. Same invisible watermark on both, if anybody checked. Everything that separated them was care – and care is invisible to every detector being built.

* * *

Which is why the new disclosure regime is so strange once you step back from it.

Substack – of all places, the home of the independent writer – now runs an optional AI-detection scan on posts and shows a “how I make this” statement beside the score, a little box where the author explains their process. I keep staring at that box, because I cannot find the precedent for it. No creator, in any medium, at any point, has been expected to file a methods statement alongside the work. I’ve watched making-of features where the actor was the only real object on set – the floor green, the ceiling green, the coffee cup green – and when the movie plays, no warning slides up to tell you the room was fake. We just call that filmmaking.

Publishing goes further. There’s an entire respectable industry – agencies, rate cards, invoices – in which somebody else writes the whole book and the person who paid gets their name on the cover. Memoirs, business books, political autobiographies. The ghostwriter signs away the credit as a term of the deal, the reader is never told, and the cover says “by” somebody who may never have opened the document. We don’t scan those books. We don’t score them. We put them on bestseller lists and invite the “author” on television to discuss their writing process. A hundred years of readers believing one person wrote what a different person actually wrote, and nobody has ever aimed a rifle at the memoir aisle. But tighten your own paragraphs with AI, under your own name, and there’s a box to fill out and a score beside your byline.

Books even have a disclosure page already. It’s called the acknowledgments: “to my editor, who made this book what it is” – a thank-you nobody audits, cheerfully papering over killed characters and rewritten endings, and here is the important part: the author decides who makes the cut. Disclosure has existed forever, as gratitude.

This summer it got reinvented as an alibi. Watch the ladder we’re climbing. First the acknowledgment: voluntary, and yours. Then the how-I-make-this box: invited, expected, increasingly assumed. Then the watermark: compelled and invisible. Then the button: not yours at all – someone else’s, aimed at you. Each rung less voluntary than the last. By the bottom rung the work no longer gets read first. The writer gets authenticated first.

The watermark deserves its own paragraph, because most people don’t know this is happening. AI companies are beginning to mark generated content, and more are likely to follow with versions of their own. In text, a mark can be woven into word choices, invisible to the reader and detectable through a system the company controls. The more words the machine chooses, the more room there is for the mark. Ghost-write with it and the signal comes through loud; proofread your own page and there may be too little to even register. The stamp measures dose, not care. It can tell you the machine was in the room, never whether anyone was minding the work. Even the FDA, the most label-happy agency we have, only makes you list ingredients. There is no law requiring soup to declare “a food processor was used in the making of this soup.” We built the equipment label anyway, and the town reads it as a contamination warning.

And look – if you personally prefer work made without AI, that is a completely legitimate taste. Like vinyl. Like hand-thrown mugs. Enjoy it proudly, seek it out, pay extra for it. Just buy it as a preference. Don’t launder it through the word “slop.”

* * *

We have run this play before, recently. From 2020 to 2022, suspicion became a label, the label became a button, the button became a consequence, and “don’t read it, report it” became a civic virtue. Both political tribes ran the same move – “fake news” was the identical machine with the polarity reversed – and the biggest platform on earth banned posts calling the virus man-made in February of 2021, then un-banned them that May when official investigations reopened the question. The virus’s origin didn’t change in those three months. The classification did. But say this much for that era: misinformation at least claimed something about the world. It could, in principle, be checked. “Seems like AI slop” doesn’t accuse you of being wrong about anything. It accuses you of a vibe.

We rebuilt the reporting machine and removed the part where something could be true or false.

And yes, before the comments write themselves: some of what’s washing up on the beach really is slop, by my definition and probably by yours. But take honest inventory of what one summer built – a button, a scanner, an invisible mark, a demerit, and reduced reach, all built around a label whose meaning is still up for grabs. Whatever the reasons – fear, taste, sheer exhaustion with the pace of change – a moral panic with a launch date is still a moral panic.

* * *

Substack’s CEO, to his credit, wrote the most honest sentence of the entire panic: “Not everything made with AI is slop, and not all slop is made with AI.” His worry is the reader who believed a person wrote something when no person did. Fair enough – but that is a disclosure question, and disclosure on the writer’s own terms is as old as liner notes. Offered, it’s liner notes. Demanded, it’s a deposition.

Because notice: this story has two buttons in it, and the town is staring at the wrong one. Slop never happens at the report button. It happens at the publish button – and behind every publish button, even the ones wired to a bot posting on a schedule, stands a human who decided no further care would be required. Somebody built that pipeline, aimed it at the feed, and walked away. The walking away is the slop. My definition says slop is what you get when the writer skips judgment. The report button lets the reader skip it too – outsourcing the verdict to a click and a piece of software. The entire regime built to punish absent judgment runs on absent judgment. Same disease. Both sides of the beach.

Here’s what history says happens next. Auto-Tune was a scandal until it was a genre – the same people reporting em dashes today will defend T-Pain to the death. Photoshop was cheating until it became a verb. The darkroom, the drum machine, the word processor: every tool panic ends the same way, with the tool going invisible and everybody back in the water. This panic will pass like all the others. Summers do – even the ones that pack the beach right up to the last day.

What we built to hunt it won’t. The buttons will still be mounted. The marks will still be sitting silently in the text years from now, readable to whoever holds the key – a fossil record of one summer’s fear. After Jaws, the men who made it watched shark-fishing tournaments boom and came to believe they’d helped load the harpoons – Spielberg has said he regrets it, and Benchley spent his final decades campaigning for the animals he’d made monstrous. The fear faded. The dead sharks stayed dead. We’re building the version of that you can’t undo with an apology.

But remember who actually solves it in the movie. Brody hates the water. That’s the joke of him – a police chief afraid of boats, marooned on an island. And when the bounties and the dock photos and the beach patrols have all failed, the answer isn’t one more report filed from the sand. It’s the guy who gets in the water anyway and does the work himself, all the way to the end. Everybody remembers the famous line about needing a bigger boat, and everybody remembers it as a line about the shark. It’s really a line about the response. When the thing in the water gets bigger, the answer is not a louder beach. It’s more nerve, more judgment, more care.

So let them mount the buttons and stamp the marks and write us up. I’m not going to sit on the sand reporting fins. I’m going to keep choosing my tools, doing the work, and proudly putting my name on what I create. Read it, argue with it, tell me where it falls short. Great – that’s a conversation worth having, and I’ll be right there having it with you.

Spielberg made Jaws with a broken machine named Bruce. But what mattered was that he knew what to show, what to leave out, and how to make the audience feel it.

See you in the water.

Subscribe now

Tomorrow, America turns 250.

The temptation on a milestone like this is to look backward – to treat two and a half centuries as a monument, finished and behind glass. That has never been what this country is. The scaffolding has never come down.

Two hundred fifty years of America is one long build. Every generation inherited something unfinished, decided the next impossible thing was worth attempting, and left the country taller than they found it. Farms became factories. Factories became rockets. Fifty years after we last walked on the Moon, we went back. And the tools always end up in the same hands – the hands of people who want to build.

There will be a lot of flags flying tomorrow, and I’ll be flying mine. But flags are easy. Steel is heavy. Pride is not denial – it is commitment, and commitment looks like building.

So that is how I am spending year 250: building. Two new ventures, hundreds of early mornings, work worth handing forward. It is the most American act there is, and it is open to every one of us, in any field, starting any morning we decide.

Two hundred fifty years in, and the best of this country has not yet been built.

There is a phrase for loving an unfinished country enough to keep building it: American Pride.

So, Happy Independence Day, America – light every deserved firework. We built 250 years of reasons to celebrate.

Share

Subscribe now

Been a while since the Vaporware Gazette made an appearance. Spent the weekend deep in a generative video project and was enjoying myself – right up until around the fortieth render of the same shot, when I started laughing at how familiar the whole ritual had become. If you’ve spent real time with these tools, you’ll recognize the moment.

#Praymaxxing #PromptlanticCity #VaporwareGazette #OccamsEraser #TechSatire

Share

Subscribe now

Artwell.ai

I’d had the kind of week with Artwell where you either ship something or break something, and I’d managed both, occasionally to the same feature. I’d also spent it neck-deep in the debate eating every feed I follow lately: the one about AI and taste, about whether, now that the machine can make the thing, having good judgment is enough to make you a creator – or only a critic. So on Friday, fried on both counts, I took myself out for a steak.

Two tables over, a couple was having an argument, and it took me a full minute to work out what they were arguing about.

She’d said the thing was technically flawless and completely forgettable. He said she was being impossible – it was obviously high quality, look at it. She said quality wasn’t the same as good. He said that was a meaningless distinction. She said it was the only distinction that mattered. He said, and I’m quoting, “everybody’s suddenly an expert.”

It was, almost word for word, the argument I’d been reading all week. So I assumed I was eavesdropping on two people from my own corner of the world – that the debate had followed me out of my screen and into dinner. I was, I admit, a little impressed by their command of the material.

Then he delivered the closer: “If you think it’s so easy, I’d love to watch you cook it better.”

Ah. The steak. They were arguing about the steak.

Here’s the part that stayed with me, past the steak and past the secondhand discomfort of watching a date go sideways over doneness. I hadn’t misheard them. I’d heard them perfectly. The argument about the steak and the argument about AI are, line for line, the same argument. You can swap “the steak” for “the output” in every sentence either of them said and lose nothing at all. That isn’t a coincidence, and it isn’t because steak is secretly profound. It’s because both arguments are circling the same confusion, and neither side has bothered to name it. But I will try.

There are two different things we call “taste,” and they only look like one thing from a distance.

The first is the diner’s taste. You take a bite and you know. Too salty, the texture’s gone rubbery, the sauce is fighting the meat instead of helping it. You don’t need a chef’s jacket to register any of that, and the man’s “I’d love to watch you cook it better” was a bad move precisely because she never claimed she could cook it. She claimed she could tell it was gray in the middle. Those are not the same claim. She was right, and he reached for a credential to avoid admitting it. Anyone who has ever been handed “well, you’re not a chef” while pointing at an obviously overcooked steak knows the move. It’s a dodge. The diner’s verdict stands on its own.

The second is the chef’s taste, and it only shows itself when the plate is wrong and has to be fixed. The diner says “this is too salty.” The chef tastes the same bite and knows the saltiness is actually a fat problem, and the fix isn’t less salt, it’s a hit of acid to cut it. The diner can reject. The chef can correct. The distance between “this is off” and “here is the specific move that makes it right” is the entire job, and it’s the part you only get from having stood at the stove and ruined the dish a few hundred times.

Hold onto that gap, because the AI conversation lives or dies in it.

The loudest story in AI right now is a democratization story, and the optimistic version of it is genuinely true. The barrier to making things has collapsed. Anyone can generate a draft, a design, a working prototype, a passable first cut of almost anything, in the time it used to take to schedule the kickoff meeting. From there the discourse has converged, with remarkable speed, on a single comforting conclusion: since the machine now handles the making, the scarce and valuable thing left to humans is taste. Judgment is the new moat. Curators beat creators. The editor outranks the engineer. There are now four-part courses promising to train your taste in ninety days, as if taste were a certification.

And the first half of that is correct. When generation is free, discernment is what’s scarce, and the person who can tell the good output from the plausible one is holding something real. The couple at the next table had stumbled into the true half: you don’t need to be a chef to know the steak is overcooked.

But watch the move the discourse makes next, because it’s the same move the man made, just run in reverse. He took “you can’t judge it unless you can cook it” – false – and used it to win an argument he was losing. The AI conversation takes “you don’t need to be a chef to judge it” – true – and quietly extends it into “so you don’t need to be a chef to make it.” That sequel does not follow. And the reason it doesn’t follow is that when you build with AI, you are no longer the diner. You’re the one in the kitchen.

Think about what the machine actually is in this arrangement. It’s a line cook with flawless technique and no palate of its own. It will execute anything you can specify, instantly, without complaint, and it will re-plate the identical wrong dish all night because it cannot taste that the dish is wrong. At a restaurant, the diner’s “this is off” is enough, because there’s a chef back there with the second kind of taste who translates the complaint into a fix. In an AI build, there is no chef back there. There’s just the line cook and you. If all you’ve brought is the diner’s palate, you can tell the plate is wrong, but you can’t say acid-not-salt, so you send it back on a loop and get something differently wrong every time. The only thing that breaks the loop is the correction, and the correction is the chef’s.

There’s a reason the whole culture reached for “slop” to name what’s wrong with mediocre machine output – a feed-trough word the dictionaries crowned word of the year. We walked all the way into the kitchen to find the insult and didn’t notice we’d done it.

Here’s what that looks like from inside my own build, on a perfectly ordinary day.

Artwell turns long-form audio into clean, speaker-attributed transcripts and then into story. So transcription quality is foundational, and at one point I had two transcripts of the same recording sitting side by side. One of them looked obviously better. It came dressed with segments, summaries, topic tags, sentiment analysis – the whole garnish tray. The other was plainer. Word-for-word they agreed about ninety-six percent of the time, close enough that a quick look would call them equivalent and reach for the prettier plate.

The plainer one was the better transcript, and it wasn’t close. The fancy one had quietly turned “Puget Sound Energy” into “Pizza Time Energy,” and a neighborhood named SoDo into “So To.” Those aren’t formatting nitpicks. In a transcript meant to carry someone’s actual words into a published story, a confidently wrong proper noun is the gray center of the steak – the thing that looks fine under the garnish and destroys trust the moment anyone who was in the room reads it. Choosing correctly here was not “which output has more features.” It was knowing that for this product, quality means the words and the names survive intact, and the analytics layer is parsley. That’s not a verdict a diner’s palate can deliver. It’s a directorial call about what the dish is even for.

It happens at every layer. Another time, a fix to the audio-editing playback sailed through every check – tests green, types clean, lint quiet – and was still wrong, because it had bought its passing grade by making playback hesitate for a beat before it started, trading the immediate, responsive feel of the thing for technical correctness. Every box ticked. The dish still didn’t eat right. Passing the recipe is not the same as the plate being good, and no test suite has ever been able to taste the difference.

Now the uncomfortable part, the part those train-your-taste courses are selling around. The chef’s palate – the one that can direct, not just judge – is not built by curating. You do not develop it by assembling a reference library of great plates and studying them, any more than you become a winemaker by becoming a sommelier. The sommelier develops exquisite preferences by tasting. The winemaker develops judgment by making bad wine for years until it stops being bad. They are different organs, grown by different work, and only one of them can fix the batch. The judgment that lets me see past “Pizza Time Energy” came from two decades of being responsible for the finished thing – of shipping the wrong cut, watching it land badly, and learning in my hands what the fix actually was. It did not arrive on a schedule, and it did not come from watching.

Which is the quiet problem underneath the cheerful democratization story. We have declared judgment the most valuable thing in the economy at the precise moment we’re dismantling the work that produces it. The reps that used to build the chef’s palate – the grunt passes, the owned mistakes, the years on the line – are exactly the reps the machine now does for you. You can have the diner’s palate for free. The chef’s palate was never free, and it’s getting more expensive to acquire, not less, because the path to it is quietly closing.

I can’t reopen the path for everyone coming up behind me, and I’m not going to pretend a LinkedIn post does anything about it. What I can control is whether I keep doing my own reps. So here’s the thing I actually do, and it’s the reason my build process looks heavier than a solo founder’s “should.” The multi-session, multi-role orchestration I run – the planning panels, the forced checkpoints, the reviewer pass that exists only to argue with the implementer pass – is not only quality control. It’s a structure that keeps me at the stove. Left alone with a fluent line cook, the gravitational pull is to slide back into the diner’s chair: glance at the plate, looks plausible, ship it, send it back if someone complains. The orchestration is what stops the slide. It forces the second-order question every time – not “is this good,” but “what is this allowed to become, and where does it live, and what’s the one cut that keeps the product true.” It makes me direct instead of merely approve. In a world where the reps don’t happen for free anymore, I’ve built a way to keep doing them on purpose. It’s an apprenticeship for one, and I’m both the master and the kid sweeping the floor.

The couple never resolved it, for the record. He paid, stiffly. She was still right.

But the argument they were having is the one the whole industry is having, and it’s worth getting right, because the stakes aren’t really about steak or even about taste. The people who can build something good with AI are the ones who already spent the years at the stove – who carry the chef’s palate into a kitchen that now runs at impossible speed. The discourse is handing everyone else a flawless line cook and a diner’s palate and calling them chefs. And that works beautifully, right up until the plate comes back wrong and nobody at the table knows what to fix.

#Artwell #FieldNotesFromTheBuild #AIStrategy #SoloFounder

Share

Subscribe now

Artwell

For the last year, most of what I’ve written has been about how companies are struggling with narrative. Meta cloning Zuckerberg into an AI nobody can shut off. Microsoft bolting Copilot onto its products, and then bolting a narrative onto Copilot to match. Big tech thought leadership that says nothing in three thousand words. AI rollouts that stall because the engineering shipped but the story didn’t.

One pattern keeps surfacing inside all of it: companies look at consumer brand giants, pick up the wrong lessons, and apply them in places those lessons don’t fit. The case study gets framed. The headline gets repeated. The slogan version of the lesson spreads. And the underlying reality of what the company actually built gets lost in the retelling.

The Golden State Warriors are one of those examples right now.

In 2022, the team launched Golden State Entertainment – a content company that produces documentaries, signs musical artists, and stages live events. Three years in, GSE has released an album featuring E-40, Too $hort, Saweetie, and Larry June; been part of an Emmy-winning short film about Jeremy Lin; and produced documentaries about figures like Mahmoud Abdul-Rauf. It gets cited often in marketing circles as a storytelling success story – proof that companies should invest in narrative infrastructure, build community, treat content as a strategic priority.

Some of that is true. But it misses what GSE actually is.

Listen to how David Kelly, the franchise’s Chief Business Officer who runs GSE, describes the venture in his own words:

“We want to leverage the team’s brand outside of sports, telling the stories of people that shape culture and move culture.”

“We want to tell some Warriors stories... but we’re not limiting ourselves to Warriors stories.”

“This is an opportunity for musicians, entertainers, creatives, and brands to partner with the Golden State Warriors to bring their projects to life.”

That’s not a sports franchise telling its own story better. That’s a media company that happens to be owned by a basketball team, using the Warriors brand as gravitational pull to attract creators who want to be associated with it. The stories being told are mostly not Warriors stories. The artists being signed are not Warriors artists. The documentaries are not Warriors documentaries. The team’s brand is the platform; other people’s work is what runs on it.

Imagine if Adobe spun up a record label, signed Bay Area musicians, and produced a slate of indie documentaries. Nobody would frame it as a marketing lesson for the SaaS industry. We’d say: Adobe is in the media business now. We’d be curious about the strategy. But we wouldn’t pretend it teaches the average B2B software company how to tell its own story better.

That’s the move GSE represents. And it’s a move that requires something most companies don’t have: cultural gravity. A brand that creators actively want to be attached to. Decades of accumulated equity. A built-in venue, audience, and pipeline. The Warriors aren’t telling stories better than everyone else. They built a media platform that lets other people tell stories under their flag, because the flag has enough pull to make it work.

This is the part that matters for most companies looking at GSE and trying to extract a storytelling lesson: they’re looking at the wrong category. The lesson isn’t “build narrative infrastructure” or “treat content as strategy.” It’s “decide whether you have enough cultural pull to be a platform – and if you don’t, which is most companies, your job is still the harder one.” Which is telling your own story, well, about your own work, to people who need to understand it. No record label is going to save you from that.

Which brings me back to the work I’ve been writing about all year. The companies struggling with narrative right now aren’t struggling because they need a record label. They’re struggling because they treat their own story as a deliverable instead of a strategy – something produced after the engineering is done, by the team three floors down. The fix isn’t to mimic the Warriors. The fix is to do the thing the Warriors are also doing, just at a different scale: name storytelling as central to what the company is, and resource it like you mean it.

Build a media company if you want one. Just don’t confuse it with the work of telling your own story.

#Artwell #AIStrategy #SoloFounder

Share

Subscribe now

Artwell

Just like Jason Voorhees, the agile-versus-waterfall debate never dies – it keeps coming back. It comes back because it’s a useful argument, and because every new wave of developers eventually has to reach their own conclusions about how much planning is enough and how much is too much. The current resurgence has a particular flavor – it lives mostly on LinkedIn and YouTube, and it leans hard against specs. Specs are heavy. Specs are corporate. Nobody reads them. Just ship. Waterfall is dead. The dinosaurs of the old way are still scolding us from the boardroom while the rest of us iterate.

There is something true in that critique, and there is something the critique is missing. Both halves are worth taking seriously, because building a real product in 2026 means navigating both at once.

The part that’s true is straightforward. The dead corporate version of a spec – the thousand-page document nobody reads, the approval chains, the false certainty, the planning that calcifies into fiction – really is a waste. That’s not the same thing as saying specs are a waste. It’s saying that one particular implementation of a spec is a waste, and confusing the implementation with the concept is the kind of category error that lets people argue against object-oriented programming because they once saw a bad Java codebase. The bad version of anything is always bad. The interesting question is whether the good version still has work to do.

The bigger problem with the current debate is that it’s being conducted in 2026 with the same binary it had in 2014, which in AI years is roughly the Cretaceous. The debate certainly is not extinct, but the framing of it is. Both sides of the argument still assume that a spec is a document a human writes for another human to read. Even when the spec gets handed to an AI rather than a human, the assumption underneath stays the same – the spec is a static artifact passed across a boundary, used as a one-time briefing rather than as a living substrate the work persists within. From that premise, the entire shape of the debate makes sense. Specs are heavy because humans are slow readers. Specs feel like waste because the people they’re written for never finish them. Defending specs makes sense from inside that frame. So does dismissing them. The two sides are both correct, and they’re both arguing about the wrong thing.

Before getting to what I actually think the debate is missing, there’s a useful principle from the older argument worth keeping, because it dissolves the binary on its own. Some decisions in front of you are expensive to reverse. Some are cheap. Architecture is expensive. A button label is cheap. You don’t iterate your way into a data model; you iterate from one. The cost of being wrong on the foundation is high, and the cost of thinking carefully about it up front is low. That phase rewards real specs. The wiring phase, where you’re shaping what the user sees and how a flow feels and where the copy lands, is different. Those things only reveal themselves once people start using them, and heavy specs there would actively hurt the work. The principle is simple: match the process weight to the reversibility of the decision. Spec the expensive, iterate the cheap.

Most serious builders already work some version of that, whether they call it that or not. The agile-versus-waterfall debate falls apart the moment you accept that different decisions in the same project deserve different amounts of upfront thinking. But the principle still isn’t the deepest thing the debate is missing. The deepest thing is that the audience for the spec has changed, and nobody on either side of the public argument seems to have noticed.

I’m building Artwell, an AI narrative and voice platform, as a solo founder. The foundational specification I worked through with multiple AI models, across many days of conversation, came out around 80 pages. The user-experience specification that followed is roughly 20 pages. I have never read either of them cover to cover, and I never will. Neither would anyone else. That isn’t the failure of those specs. That is the point of those specs.

When you build a real product with AI in the loop – through model versions that change, across context windows that overflow and crash, across sessions you start fresh tomorrow because the current one is exhausted, across multiple models running in different roles – you need somewhere for the work to live that isn’t just your head and isn’t just one fragile session. The spec is that somewhere. A new session, a new model, a new tool reads the document in seconds and comes up to speed on weeks of accumulated decisions, definitions, constraints, edge cases, and intentions. Without that document, every restart is a regression and every model swap is a tax. With it, the work persists across all of it.

And the conversation that produces the spec is itself iterative, which is the part that breaks the frame entirely. The 80-page foundational spec wasn’t a waterfall artifact handed down from a planning phase. It was the consolidated output of multi-day back-and-forths across multiple sessions, multiple models, and multiple roles – a planner session proposing, a reviewer session pushing back, an implementer session surfacing a hole the planner missed. Features I would never have discovered through pure vibe-coding came out of the planning conversation itself, because that conversation had the time and the surface area to let edge cases and competitor analysis and full workflows breathe. The output happens to be a long document, but the document is the consolidated record of an extremely agile process. The thing the public debate calls “spec writing” and the thing it calls “iteration” are happening in the same place. They just produce different artifacts at different points.

That changes the cost calculation entirely. A 50-page spec that humans have to read end-to-end before they can start working is corporate theater, and the critics are right to dismiss it. A 50-page spec that an AI ingests in seconds and that survives the next model swap is infrastructure, and dismissing it is a category error. Same document. Completely different function. Whether the work of writing it is wasted depends entirely on who’s reading it and what they’re using it for.

And the spec itself doesn’t stop evolving once implementation begins. We treat it as the working substrate of the project – not just as a planning artifact but as the implementation document itself, broken into phases, lanes, and slices that get worked through over the following weeks and months. As that work surfaces things the planning didn’t anticipate, the spec gets updated. Sometimes we revise the original document directly. Sometimes we preserve the original intact for the historical record and write addendums that capture founder-approved course corrections. Either way, the spec stays current with the actual state of the project, which means new sessions and new models always get the truth of where we are – not a snapshot of where we thought we’d be six weeks ago. The flexibility to change direction during implementation doesn’t break the foundation. It’s built on top of it, which is the entire reason the foundation was worth specifying in the first place.

This isn’t new thinking outside of software. Many years in media production taught me that the work you put into pre-production is what makes production possible. An indie filmmaker can shoot from the hip with no script, no shot list, no production schedule, and pull off something charming – the kind of thing that wins a festival short slot and becomes a calling card. The same filmmaker cannot pull off a feature that way. They certainly cannot pull off a series. The amount of structure required scales with the ambition of the thing being built. The script that nobody reads cover-to-cover on set is doing real work as a shared reference. The shot list the DP scribbles all over is doing real work as a daily handoff. The shooting schedule that gets revised every morning is doing real work as a coordination layer. None of it is waterfall, and none of it is wasted. It’s the substrate that makes the days on set possible.

Specs work the same way. A weekend script doesn’t need one. A product with customers, retention policies, billing, trademarks, compliance posture, and a brand audience whose trust breaks if anything breaks does. What changed isn’t whether the substrate matters. What changed is who reads it.

There’s a related point that follows from this, and it matters because it’s where the public debate quietly does the most damage. When people complain that AI claimed it finished a task it didn’t actually finish, or that the AI skipped a section of the spec, or that the AI ignored a constraint, the default response is to blame the AI. The more honest diagnosis is that the AI is doing exactly what its scaffolding allows it to do. A single prompt-and-pray session with no checkpoint, no completion verification, no review panel, and no orchestration layer is going to produce inconsistent results regardless of which model is on the other end. That isn’t an AI failure. That’s a builder choosing not to construct the orchestration that the work requires.

The builders who are shipping serious products with AI assistance are running multi-session, multi-model, multi-role workflows with forced checkpoints, completion verification, and review passes. The builders who are shipping weekend tools and clever demos can get away with one session and a good prompt. Both are doing AI-assisted development. They are not doing the same job. The advice that works for the second does not port to the first, and treating the two as equivalent is the original sin underneath most of the loudest takes on agile, specs, and how to build with AI today.

So the real debate isn’t agile versus waterfall, and it isn’t even spec versus no-spec. It’s whether you’re building something that requires a real orchestration architecture and a persistent record of its own thinking, or whether you’re building something that doesn’t. It’s whether the documents you’re producing are written for human readers who’ll never finish them, or for the AI collaborators who actually carry the work across the sessions you can’t.

I don’t read my own specs. They aren’t for me. They’re for the next session, the next model, the next implementer, the next reviewer, the next time the context window blows up and tomorrow’s version of the project has to be reconstituted from yesterday’s decisions. The specs are how the project remembers itself. They are the foundation that makes everything else I do possible, even though they look, on the surface, like exactly the kind of artifact the current debate has decided to dismiss.

That isn’t waterfall coming back. It isn’t the old discipline in new clothes. It’s a new discipline that the public debate hasn’t built the vocabulary for yet – and one that the people still arguing about specs versus no-specs are going to need, whether they decide to learn it or not.

#Artwell #FieldNotesFromTheBuild #AIStrategy #SoloFounder

Share

Subscribe now

Artwell

Nine months ago, when I began building Artwell in earnest, Deepgram was the right call for speech-to-text. Best quality at the time. Best pricing. Solid intelligence layer.

I also flagged, in the first week, that Artwell should not be permanently coupled to a single transcription provider. That was not some brilliant revelation. Every senior architect would have seen the same thing.

But there is a difference between seeing an architectural need and building it immediately. Early in the build, the more important job was learning the shape of the product: the workflow, the transcript model, the editing layer, the downstream story outputs, the places where quality actually mattered. Premature abstraction is its own kind of debt. So we kept moving.

This week I went back and re-tested the speech-to-text market against Artwell’s real needs – long-form audio, diarization, speaker handling, intelligence layers, downstream editing, cost at scale. The landscape had not moved by inches. It had moved by miles.

ElevenLabs’ latest Scribe v2 has overtaken on quality and pricing. AssemblyAI is right behind it, with a comparable intelligence layer – sentiment, topics, summaries, speaker identification inferred from transcript context. For the configuration Artwell needs, the incumbent provider now costs more than double what the leaders charge, with measurably higher word error rates (WER) and a thinner intelligence layer. This is no small thing. This changes margins, pricing flexibility, and what a product can afford to offer users.

With Artwell’s launch on the horizon, this became the moment to build the abstraction layer.

Artwell now has a transcription layer that can route across providers, support fallbacks, separate development from production, and absorb the next provider change without touching the product itself. The intelligence layer is next, same pattern – OpenAI and Anthropic today, whatever proves best tomorrow.

The lesson though is not “build abstraction layers.” The lesson is that AI infrastructure decisions have a shorter shelf life than most teams are conditioned for. The right provider today may not be the right provider one year from now. It might not even be the right provider one month from now. That’s simply the speed of the market we are all building in.

“Which provider is best” is the question every team asks. The better question to ask is what it will cost you to switch when that answer changes.

Architecture buys you the right to change your mind.

And in AI, that optionality may be the difference between building a moat and building a wrapper.

𝘛𝘩𝘢𝘵 𝘪𝘴 𝘸𝘩𝘢𝘵 𝘈𝘐 𝘴𝘵𝘳𝘢𝘵𝘦𝘨𝘺 𝘭𝘰𝘰𝘬𝘴 𝘭𝘪𝘬𝘦 𝘧𝘳𝘰𝘮 𝘛𝘩𝘦 𝘉𝘶𝘪𝘭𝘥.

#Artwell #FieldNotesFromTheBuild #AIStrategy #SoloFounder

Share

Subscribe now

Artwell

Two stories get cited every time someone recites the lean MVP gospel: WhatsApp and Craigslist. Both are real. Neither is actually about the product.

WhatsApp launched in 2009 with a Russian contractor hired off RentACoder, and the first version crashed so often that founder Jan Koum almost gave up and went job hunting. What saved it wasn’t iteration on user feedback – it was Apple later shipping push notifications, and Koum knowing exactly what to do with them. The founders weren’t kids racing to ship. They were two former Yahoo engineers in their forties with twenty years of combined experience, building in a market where “free international messaging” was a structural advantage against expensive SMS. By the time Facebook bought it for $19 billion in 2014, the company still had only 56 employees. The product was thin. The founders weren’t.

Craigslist is even less of a “ship fast” story. Craig Newmark was a 40-something engineer with seventeen years at IBM when he started emailing event listings to friends in 1995. The site didn’t become a company until 1999. He treated it as a side project for four years while his day job was contract programming. The competition was newspaper classifieds. There was no saturation, no race, no MVP gospel – just an experienced engineer letting a network effect build quietly because he understood, structurally, what he was actually building.

Strip the retrofitted “vanilla MVP, iterate fast” narrative away and the same pattern surfaces in both stories. Experienced operators. Structural market reads. Tools so primitive that “minimum” wasn’t a strategy, it was the only option. The product wasn’t the differentiator. The founder was. We just didn’t notice because the tools obscured it.

In 2026, the tools have commoditized, and the founder is the signal.

Recently I spent four days writing and rewriting a 20-page UX architecture document for Artwell – workflow definitions, workspace logic, interaction rules, edge cases, language, and intent. Then I handed it to Claude’s new design tool. Within an hour, I had beautiful, meaningful, production-quality interface templates. But the real story isn’t that one hour – it’s those four days. The tool turned judgment into pixels at speed. Without the judgment, it would have produced beautifully designed nothing.

That’s the shift the MVP discourse hasn’t caught up to yet. When tooling was primitive, the product was the bottleneck, so the product was the variable. You shipped vanilla because you couldn’t ship anything else, and you iterated because the market was unsaturated enough to give you the runway. In 2026, none of that is true. Tools are commoditized. Markets are saturated. Any serious founder can ship something competent in a fraction of the time it used to take. So what’s left to differentiate?

The founder. That’s it. That’s the only variable that doesn’t commoditize.

Which is why some of the loudest MVP advice online feels increasingly incomplete. It isn’t just vibe coding anymore. It’s vibe architecting, vibe marketing, vibe founding. AI is being used to compensate for judgment that hasn’t been built yet, and the output reflects exactly that. Beautifully designed nothing, shipped fast, drowned in a saturated market full of other beautifully designed nothings. The amplifier did its job. The signal underneath was the problem.

The MVP framework wasn’t wrong; it was a product of its constraints, and those constraints have changed. The new minimum bar isn’t a feature set. It’s a founder. A minimum viable founder – someone with enough judgment, taste, market read, and conviction to know what to build before they touch a tool, and to know why it matters before they ask anyone else to care.

If you’re building right now, that’s the work. Not the stack. Not the speed. Not the iteration cadence. Invest in your read of the market. Invest in your editorial taste. Invest in your ability to write a comprehensive UX document that tells a tool exactly what to make and why. The product will follow. Tools are too good now for it not to.

The founder is the differentiator. The tools didn’t change that. They just made it impossible to miss.

#Artwell #AIStrategy #SoloFounder

Share

Subscribe now

Artwell

A few months ago I wrote that American pride starts with remembering who we are – and that a country which stops doing hard things eventually forgets it ever could.

Today, Artemis II came home.

Four astronauts, a lunar flyby, and a splashdown that looked a lot like the photos my parents watched live on television more than fifty years ago. Except this wasn't a memory. This was hours ago.

I have been thinking all evening about what it felt like to watch it. Not the engineering, though the engineering is staggering. Not the politics, because this moment belongs to something larger than that. What I felt was the thing I wrote about and was not sure I would feel again in my lifetime: the quiet, uncomplicated pride of watching my country do something unambiguously worth being proud of.

Pride is not denial. It is commitment. It is the willingness to believe that building impossible things together is still who we are, and then going and proving it.

Today we proved it. Welcome home, Artemis II. And thank you for reminding a lot of us how that's supposed to feel.

#Artwell #AIStrategy #SoloFounder #ArtemisII

Share

Subscribe now

Artwell

Breakthrough: Scientists have successfully replicated Schrödinger's quantum superposition using only venture capital and high confidence.

#SchrödiconValley #QuantumUnicorn #VaporwareGazette #OccamsEraser #TechSatire

Share

Subscribe now

Artwell

When I was a kid I had a black-and-white photo of Bruce Lee pinned to my bedroom mirror. I was fat. I got picked on. And every morning I woke up and looked at a man who the world underestimated until they couldn’t anymore.

I didn’t care about the kicks. I cared about the philosophy.

Bruce Lee studied every martial art he could find, and then he threw them all away. Not because they were worthless. Because they had become more important than the fight itself. The forms were beautiful. The katas were precise. And they would get you killed, because they trained you to follow a pattern instead of responding to what was actually in front of you.

But here’s what most people leave out of that story. He didn’t walk away from nothing. He mastered Wing Chun under Ip Man. He studied Western boxing, fencing, judo, wrestling, grappling, taekwondo. He didn’t reject discipline – he completed it. Jeet Kune Do wasn’t the absence of form. It was what came after mastery of form. The formless form. And that distinction matters, because anyone can tear up a playbook. Only someone who has mastered the playbook earns the right to replace it with something better.

He replaced all of it with one principle. He inscribed it on the Jeet Kune Do symbol in Chinese characters:

Using no way as way. Having no limitation as limitation.

The whole world remembers “be like water.” Almost nobody remembers what he actually meant. He wasn’t talking about style. Everybody gets that wrong. He was talking about form. Rigid form. He was warning against it. The kind of form that looks powerful in a controlled environment and shatters the second reality shows up unannounced.

I’ve been thinking about this a lot lately. Because the AI industry is building nothing but form.

The Industry Is Performing Kata

Every major company deploying AI right now is performing kata.

Choreographed sequences. Executed the same way every time. There’s a transformation roadmap. A change management framework. An all-hands where a senior leader says the words “more productive.” A polished internal memo. An external campaign.

Beautiful choreography.

Adoption goes up 13%. Trust goes down 18%. 77% of employees say AI increased their workload. 56% of companies report zero financial return. 59% of hiring managers admit they dress layoff announcements in AI language because it plays better than telling the truth.

I Watched This Happen From the Inside

I spent almost twenty years inside Microsoft. I watched this happen in real time. Not once. Over and over. The playbook kept running even after reality had moved on. The structure was still standing, the choreography was still being performed, but the truth had left the building and nobody noticed because the form was still intact.

Bruce Lee saw this exact pattern in every traditional martial art that had stopped questioning itself:

“Set patterns, incapable of adaptability, of pliability, only offer a better cage. Truth is outside of all patterns.”

– Bruce Lee

The cage is the playbook. The cage is the framework. The cage is the bolted-on narrative that tells employees AI will empower them while their colleagues disappear.

The Imitation Problem

There’s a quote from the Tao of Jeet Kune Do that I forgot for thirty years and remembered tonight.

“We have more faith in what we imitate than in what we originate. We cannot derive a sense of absolute certitude from anything which has its roots in us.”

– Bruce Lee, Tao of Jeet Kune Do

Read that again in the context of how companies are deploying AI.

Every enterprise is imitating every other enterprise. The same playbooks. The same transformation language. The same announcements. The same Copilot rollout strategy. Nobody is originating anything because originating requires trusting your own judgment about what your people actually need. And that’s terrifying when the whole industry is moving at once and nobody wants to be the one who did it differently.

So they imitate. They bolt on what worked somewhere else. And they wonder why it doesn’t work here.

“If you follow the classical pattern, you are understanding the routine, the tradition, the shadow – you are not understanding yourself.”

– Bruce Lee

That’s the AI industry’s problem in one sentence. Companies are following the classical pattern. They understand the routine. They do not understand themselves. They do not understand their own people, their own culture, their own customers, the actual lived reality of the humans who are supposed to benefit from all of this.

The Audience Has Already Left

Half of Americans say AI in daily life makes them more concerned than excited. 65% of workers barely use it. The people building AI narratives live in a bubble where everyone uses AI every day, and they assume the rest of the world is right there with them.

It isn’t.

Coca-Cola’s AI holiday ads got rejected as soulless. McDonald’s pulled their AI Christmas commercial after audiences called it slop. iHeartMedia launched a “Guaranteed Human” campaign because 9 in 10 consumers want their media made by actual people. Merriam-Webster named “slop” their word of the year.

The forms keep performing. The audience has already left.

What I’d Say If I Could Be Honest

If I could sit across from every CEO pushing AI transformation and say one honest thing, here’s what I’d say.

Your version of AI is crushing something deeply human. It’s crushing endeavor. It’s crushing dignity in work. People feel it even when they can’t articulate it, and your internal communications aren’t fooling anyone.

But here’s where most critics get it wrong, and I want to be precise.

AI is not the problem.

I use AI every waking hour. It replaced Google for me. It replaced how-to videos. It replaced a hundred small friction points. It is one of the most powerful tools I have ever used. I am building an entire company on it.

The problem is how AI has been framed, marketed, and introduced to the world. We have done a terrible job helping people understand where AI helps, where it harms, and where the line should be. And we have done an even worse job listening to what people actually need before deciding what to build and how to sell it.

Not better AI. Better honesty about AI.

Hack Away at the Inessentials

Bruce Lee didn’t build Jeet Kune Do by learning more techniques. He said it himself:

“It is not a daily increase, but a daily decrease. Hack away at the inessentials.”

– Bruce Lee

The AI industry needs to hack away at the inessentials. The theater. The hype. The fake wisdom. The obsession with scale for the sake of scale. The idea that making humans smaller somehow counts as progress.

The version of AI nobody is building toward is the one that quietly makes people stronger. Smarter.Clearer. More capable. Not more dependent. Not more passive. Not more replaceable.

Something that serves the human without swallowing the human. Powerful, but disciplined. Something that helps you move better, not disappear faster.

A formless form. Not chaos. Readiness. The ability to respond to what’s real without being locked into a pattern designed for a different fight.

The Mirror

When I was thirteen I pinned Bruce Lee to my mirror because I wanted to become something better than what I was. I was a fat kid who got picked on, and he represented the possibility that the person everyone underestimates can be the one who sees most clearly.

So I did something about it. I started training in Shotokan karate. I became a Varsity wrestler. Not because I wanted to fight anybody. Because I wanted to stop being afraid. And it worked. Not overnight, not dramatically, but in the way real things work – slowly, through repetition, through showing up when I didn’t feel like it. I lost the weight. I found confidence I didn’t know I had. I learned that discipline wasn’t something you performed. It was something you practiced until it became part of you.

That kid went on to spend twenty years at Microsoft and build a content program from nothing into something that generated $59 million in revenue. I mastered the playbook. And then I earned the right to throw it away and build something of my own. And I forgot, for a very long time, where all of it started.

I’m older now. I forgot that photo. I forgot the philosophy. I let decades of corporate life and monthly bills and other people’s priorities bury something that used to matter to me more than anything.

I’m not forgetting it again.

If I could pin something to my mirror today it would say this:

You built today. You earned tomorrow.

That’s not a Bruce Lee quote. That’s mine. But I think he’d understand it. Because it’s about the same thing he spent his whole life teaching – that truth is earned through honest work, not inherited through rigid form. And that the person willing to throw away the playbook and start from what’s real will always see further than the ones still performing kata in an empty room.

Be like water.

Bruce Bracken is the founder of Artwell.ai and former Head of Podcasts & Digital Experiences at Microsoft, where he built a content ecosystem that generated $59M in attributed Azure revenue. He writes about the gap between what AI companies say and what humans actually experience.

Share

Subscribe now

Artwell

In December, I published an article making the case that Microsoft’s real AI problem was integration, not capability. Copilot wasn’t failing because the intelligence wasn’t good enough. It was failing because AI showed up in their products as something added rather than something assumed. A sidebar. A button. A helper. Never the organizing principle of the experience itself. I called it Artificial Integration.

Three months later, Microsoft announced Copilot Cowork, integrating Anthropic’s Claude Cowork technology into Microsoft 365 Copilot. The move made strategic sense. But it also reinforced the same pattern: even when a company brings in genuinely powerful AI capability, the result still feels bolted on. The platform absorbs the innovation without reorganizing around it.

But this article isn’t about products. It’s about something companies are getting even more wrong, and almost nobody is talking about it.

The Same Mistake, One Level Up

Every major technology company is currently telling two stories about AI. Internally, they’re telling employees that AI will make them more productive and more capable. Externally, they’re telling consumers that AI makes everything better. Both stories share the same structural flaw: they’re bolted on.

Companies are bolting AI narratives onto their organizations the same way Microsoft bolted Copilot onto Office. The narrative shows up as a memo, a talking point, an all-hands presentation, an ad campaign. Always disconnected from where people actually live. The story exists in a completely different psychological space than the lived experience. And people can feel the gap immediately.

This is an architecture problem, not a communications problem. And it’s the same one I wrote about in December, just one level up from the product.

The Internal Narrative Is Broken

Companies are throwing AI narratives at the wall like spaghetti, hoping something sticks. A productivity story here. A transformation memo there. An empowerment talking point in the all-hands. None of it connects to what employees actually experience when they sit down at their desk, and the data makes the disconnect impossible to ignore.

ManpowerGroup’s 2026 Global Talent Barometer found that while regular AI usage jumped 13% in 2025, confidence in the technology plummeted 18%. Adoption going up while trust goes down is a signal that the story around the technology is failing, not the technology itself.

Great Place to Work found a staggering gap between how executives and frontline employees perceive AI communication. Executives overwhelmingly believe their messaging is landing. Frontline workers largely disagree. It’s one of the widest perception divides in recent workplace research.

Upwork found that 96% of C-suite leaders expect AI to boost productivity, while 77% of employees say the tools have actually increased their workload. Leadership is telling a productivity story. Employees are living the opposite of it.

PwC’s 29th Global CEO Survey found that 56% of companies have seen zero financial return from AI investments. Zero. Only 12% reported both revenue gains and cost reductions. Hundreds of billions spent, and most companies have nothing to show for it.

And then there’s the layoff story. A December 2025 Resume.org survey found that 59% of hiring managers admit they emphasize AI in layoff announcements because it “plays better with stakeholders” than admitting financial constraints. Oxford Internet Institute researchers have documented this directly, noting that companies are scapegoating AI to disguise conventional cost-cutting.

None of this is what a real narrative strategy looks like. Strategists don’t throw spaghetti. They research their audience, they listen, they build plans around what people actually need to hear and experience. What the data shows is that most companies don’t have anyone doing that work. They have messaging without strategy, communications without architecture, and stories that no one on the receiving end believes.

The 2026 Edelman Trust Barometer confirms it. When asked who they trust to tell the truth about AI, workers chose their peers over CEOs by a factor of two. Two-thirds of AI distrusters said adoption feels forced, not voluntary. The people sending the bolt-on narratives are the people employees trust least.

The Consumer Narrative Is Broken Too

The internal story is failing. The external one might be worse.

Pew Research Center found that half of U.S. adults say AI in daily life makes them more concerned than excited, up from 37% in 2021. Only 10% are more excited than concerned. And 65% of American workers still say they don’t use AI much or at all. The people building AI narratives are talking to the 35%. The other 65% are watching from the outside, increasingly anxious about something they barely interact with.

The people building these narratives are geographically and professionally concentrated in exactly the places you’d expect. AI users skew urban, skew educated, skew toward banking, finance, and technology. The teams creating AI advertising campaigns and consumer messaging live inside a bubble where everyone uses AI every day. They assume the rest of the world is right there with them. It isn’t.

This is why we keep seeing the same failures repeat. Coca-Cola’s AI-generated holiday campaigns, two years running, were rejected as soulless. McDonald’s Netherlands pulled its AI Christmas ad after audiences called it “AI slop” that “ruined my Christmas spirit.” Researchers call this the “authenticity premium,” a measurable trust penalty that activates the moment consumers sense a machine is behind the message. A study from the Nuremberg Institute for Market Decisions found that simply labeling an ad as AI-generated makes people perceive it as less natural and less useful.

Here’s the part that should make every marketer uncomfortable. The McDonald’s AI ad required seven weeks of production with a team of AI specialists doing extensive refinement on each shot. They didn’t save time. They didn’t save money. They just lost the human element. And observers have pointed out that AI companies are now caught telling two contradictory stories at once: telling the public that AI content requires just as much human effort as traditional production, while telling investors it eliminates the need for human labor.

Compare this to what great technology storytelling actually looks like. Apple’s “Misunderstood” ad from 2013. The teenager who appears antisocial at Christmas, glued to his phone, until he reveals he’s been making a beautiful family video the whole time. The ad didn’t mention a single product feature. The technology was invisible. The story was about a family, and the product was just the tool that made the moment possible. That ad destroyed people emotionally because it started with the human experience, not the technology. The iPhone was the least important thing in the frame. The fact that one of the most recognized examples of native technology storytelling is over a decade old tells you something about how rare this actually is.

The Backlash Is Already Here

The market is responding to the broken narrative in ways that should concern anyone building an AI strategy. iHeartMedia launched a “Guaranteed Human” campaign built on research showing that 9 in 10 consumers want their media created by real people, even consumers who use AI tools themselves. iHeartMedia CEO Bob Pittman put it bluntly: consumers aren’t just looking for convenience. They’re searching for meaning.

In Hollywood, the credits of the Apple TV hit “Pluribus” from Vince Gilligan now read: “This show was made by humans.” Across New York City, subway ads for the AI wearable “Friend” have been vandalized with messages like “AI is not your friend” and “talk to a neighbor.” An artist created a browser extension called Slop Evader that filters search results to include only content from before ChatGPT’s release.

Merriam-Webster chose “slop” as their 2025 word of the year.

Gartner found that 53% of consumers distrust AI-powered search results, and 61% want the ability to toggle AI on or off. People don’t want AI removed from the world. They want control over when and how it enters their experience. That distinction matters enormously, and bolt-on narrative ignores it entirely.

The Pattern

Three months ago, I argued that Microsoft’s AI was smart enough but not native enough. That diagnosis applies far beyond products now.

Companies are treating AI adoption as a deployment problem. Deploy the technology, then communicate about it. Ship the product, then explain it. The story always arrives after the fact, as an afterthought, disconnected from where people actually experience the change.

The companies that win the next decade will be the ones that understand something the current generation of corporate leadership mostly doesn’t. Adoption is a narrative design problem, and the narrative has to be built into the experience from the beginning, not applied afterward like wallpaper over drywall.

Most organizations don’t have anyone thinking about this at a systems level. The communications function in most companies was designed to announce decisions, not to shape them. Change management still runs on playbooks written before generative AI existed. And marketing departments are producing AI-themed campaigns from inside a bubble that looks nothing like the world their customers actually inhabit. The role that should be doing this work hasn’t been built yet in most organizations.

The gap between what companies say about AI and what people actually experience keeps widening. And almost nobody in corporate leadership recognizes that the story is the adoption mechanism. Not decoration around it. Not a follow-up to the deployment plan. The mechanism itself.

I diagnosed the product version of this problem in December. The narrative version is next.

Bruce Bracken is the founder of Artwell.ai and former Head of Podcasts & Digital Experiences at Microsoft, where he built a content ecosystem that generated $59M in attributed Azure revenue. He writes about the gap between what AI companies say and what humans actually experience.

Share

Subscribe now

Artwell

The best writing rooms don’t happen in writing rooms. They happen at a bar at 11pm when five opinionated writers are three drinks in, arguing over the same draft. One catches the plot hole. One catches the bad dialogue. One catches the joke that’s technically funny but kills the pacing. Nobody gets the whole picture alone. The final script is better than any single person at the table could have produced.

That’s also the best description I’ve found for how to actually use AI well – except that’s not how most people use it. Most people are sitting at that bar alone, scribbling “make something good” on a cocktail napkin, sliding it to the bartender, and leaving a one-star review when they don’t get the drink they wanted.

I spend more time than I should on Reddit watching people declare AI models are broken. “GPT deleted my entire codebase.” “Claude refused to do what I asked.” “Gemini hallucinated a lawsuit that doesn’t exist.” And every time, buried about four comments deep, someone asks the question nobody wants to answer: “What was your prompt?”

*crickets*

There’s a phrase people love to say when they want to sound pragmatic about AI: garbage in, garbage out. It’s old. It’s unsexy. It’s also the closest thing we have to a law of physics in this space. The part that keeps surprising me is how many “AI power users” repeat that phrase while continuing to feed models prompts that look like they were typed with one thumb at a red light.

And I’m not talking about casual use. I’m talking about high-stakes work – code review gates, implementation checkpoints, security-sensitive pathways, decisions that either prevent damage… or politely approve it. Every vague prompt that produces bad code becomes tech debt you won’t discover until it bites you in production six months from now. Every contradictory instruction that forces a rerun – and then a re-rerun – burns tokens, burns hours, and burns the subscription you’re paying for whether the output is useful or not.

Over the last two weeks, I took on a full foundation refactor of Artwell, the Story Engine platform I’m building. Fourteen phases, dozens of implementation slices, fourteen days straight – the kind of rebuild where you either get the architecture right or you get to do it again in six months. Because the stakes were high and the surface area was enormous, I decided to run three frontier models as a review panel across the entire process, from the initial planning review through every meaningful checkpoint along the way. Not a vibe check. Not cherry-picked screenshots. A repeated, evidence-backed panel as the work moved from plan to implementation to remediation. Dozens of verdicts. Reruns when blockers were disputed. Retrospective tracking when reality proved who was right and who was merely confident.

What fell out of that process wasn’t just a ranking. It was a sharper view of what most “model evaluations” actually are: prompt evaluations in disguise. When the instructions were coherent, the panel looked brilliant. When the instructions contradicted themselves, even the best models began behaving like perfectly obedient interns following a bad spec – faithfully, catastrophically, and without apology.

One of the cleanest examples was also one of the most banal: a prompt pack that required mandatory reads while simultaneously enforcing contamination rules that prohibited reading them. The model didn’t hallucinate. It didn’t get confused. It did exactly what it was told to do – and then hard-stopped on a procedural blocker because the human who wrote the instructions created an impossible situation.

That should feel familiar to anyone who’s ever handed someone two conflicting priorities and then acted surprised when they froze. Tell your direct report “move fast” and “don’t break anything” without defining the boundary, and watch what happens. The only difference with AI is that the failure happens in language first – which makes it easier to dismiss as “model weirdness” instead of what it actually is: governance debt.

The Other Kind of Failure

There’s also a second prompt failure mode that’s almost the opposite of contradiction, and it’s sneakier because it produces beautiful green dashboards.

Prompts that are too strict. Too narrow. Too “paint-by-numbers.”

When the guardrails are a millimeter apart, models will stay perfectly between them, report everything is green, and never once mention the cliff right next to the track. If your prompt only asks, “Does the existing checklist pass,” the model can honestly answer “yes” while completely missing that the checklist itself failed to ask the right questions. That’s not the model being dumb – that’s the model being obedient.

I’ve seen this happen in a particularly irritating way: you can almost watch the model notice something odd during its reasoning – a strange invariant, a suspicious edge path, a “wait, why is that allowed?” moment – and then omit it from the final report because the prompt never explicitly gave permission to include “extra” observations. The model isn’t hiding it. It’s following your instructions like a watchdog you accidentally trained to wag at intruders.

And the missed defect doesn’t vanish. It ships. It hides in your codebase, passing every check you built, until some future Tuesday when it surfaces as a production incident and you’re suddenly paying emergency rates to fix something a better prompt would have caught for free.

So What’s the Right Balance?

Prompt quality isn’t just “be comprehensive.” It’s “aim for the right balance.”

Some reviews should be surgical. You want a model to prove a specific invariant and treat anything else as noise. That’s the right move when you’re validating a security promise, or locking a contract, or verifying a remediation. In those moments, freedom is the enemy and precision is the product.

Other reviews benefit from breathing room – permission to look outside the planned lane, to ask “what changed during implementation,” to flag weirdness even if it’s not on the checklist, and to challenge assumptions the planning stage might have missed. That’s how you catch the things that slip through, not because anyone was careless, but because real systems evolve in ways planning documents can’t perfectly predict.

And this is where the models come in – because each one failed in ways that illuminate a different kind of prompt problem.

The Review Panel

Across roughly thirty tracked review cycles – planning reviews, checkpoint gates, reruns, the whole messy parade – the overall effectiveness ranking came out as:

1. GPT-5.3-codex-xhigh

2. GPT-5.2-xhigh

3. Claude Opus-4.6

If you stop at the ranking, you’ll miss what matters, because the story is really about failure patterns and role-fit.

GPT-5.3 was the most reliable sentinel – fast, cross-stack realistic, and consistently first to catch the quiet defects that make a system appear to work until it doesn’t. Its weakness was the cleanest illustration of the prompt contradiction problem: give it conflicting instructions and it will hard-stop every time. You wrote an unsatisfiable contract and then blamed the interpreter.

GPT-5.2 is the deep defect hunter – the model that obsesses over control flow, schema validation, and security leakage at the level where things look fine from a distance and fail spectacularly up close. It produced some of the highest-value unique catches in the entire dataset, finding blocker-class bugs the other two models cleared without hesitation. I’ll be honest: I spent weeks hoping the data would let me drop this model from the panel just to speed things up. Its response times could run three to five times longer than the others. The data didn’t cooperate. The catches were too important.

Opus-4.6 is the strongest architecture and process synthesizer in the group – the one you want reading the whole packet and telling you whether the system makes sense as a system, not just whether a diff compiles. Where it fell short was first-pass blocker severity. It can be, in a very human way, too reasonable. It gives the benefit of the doubt. That’s a great trait in a colleague. It’s a risky trait in a gatekeeper. In several cycles, Claude said “Ready” while the other models caught real, later-validated bugs. The prompt gave it permission to pass, and it took it.

Every model had moments of clarity and brilliance. Every model had moments of complete failure. And the failures didn’t follow a simple pattern – the one that caught a critical security gap in one cycle missed a durability blocker in another.

Why the Answer Isn’t “Pick the Best One”

This is why the best approach wasn’t mono-model. It was a panel with roles.

The strongest configuration was GPT-5.3 and GPT-5.2 as the primary technical gate pair – speed and durability coverage on one side, adversarial depth on the other – with Claude as the coherence and verification seat that forces the work to make sense and confirms closure is real. Not just different prompts. Different brains. These are fundamentally different LLMs trained on different data, and part of the value is that they genuinely see different things – the way three writers at that bar will catch three different problems in the same draft because they read with different instincts.

Think of it as separation of concerns applied to judgment. In a writing room, the person who spots plot holes isn’t always the person who fixes pacing – and you don’t force them to be. Detection and validation are different cognitive tasks. You wouldn’t ask the same engineer to both find the bugs and certify the release. You shouldn’t ask the same model to do both, either.

And the prompt design for each seat was different. The sentinel seat got tight, specific instructions: prove this invariant, flag this class of defect, stop on this condition. The depth seat got wider scope: look for what the checklist missed, challenge assumptions, interrogate edge paths. The synthesis seat got a holistic mandate: does this system make sense end-to-end, and is closure real?

When I matched the prompt to the role and the role to the model’s strengths, the panel was remarkably effective. When I used the same prompt for everyone, I got noise from the fast reviewer, missed defects from the synthesizer, and procedural gridlock from the deep hunter. Same models. Different outcomes. The variable was the prompt.

The Part People Keep Skipping

And that brings me back to the part people keep skipping because it feels annoyingly “process-y”:

Writing a good prompt is not typing. It’s engineering.

It requires you to know what you actually want, resolve contradictions before you outsource reasoning, and decide whether you’re trying to validate a known checklist or invite the model to challenge the checklist itself. It forces you to define what counts as a blocker and what counts as follow-up. It exposes gaps in your own thinking because the model will walk exactly where you pointed – and if you pointed into fog, it will give you fog-shaped certainty right back.

The next year of AI won’t be won by the teams with the fanciest model access. Everyone has access now. The advantage is shifting to teams who build systems around models – workflows that treat prompting as an engineering artifact, panels that separate detection from validation, and prompt packs that don’t contain landmines or blinders.

In other words, teams who stop treating prompts like a text field and start treating them like a spec.

I started this process thinking I was evaluating models. Two weeks later, staring at thirty cycles of evidence, I realized the models had been evaluating me the whole time. Every false alarm traced back to a contradictory instruction I wrote. Every missed defect traced back to a prompt that wasn’t brave enough to ask the real question. Every time the panel disagreed, it was because I’d given three different brains permission to read the same instructions three different ways.

Garbage in, garbage out. Turns out it’s not just a law for machines.

Share

Subscribe now

Artwell

Gracie found the scent again this morning. Same patch of grass, same obsessive circling, same conviction that the deer or rabbit she smelled there three days ago might still be waiting. I’ve watched my golden retriever copilot do this a hundred times on our early morning walks – past signal, possible reward, investigate again – and it never changes. She trusts the pattern completely.

At the same time I was on my phone working through a strategic question out loud with ChatGPT. At one point, the model responded by pointing out that I was relying heavily on pattern matching to reach my conclusion.

That caught me a little off guard. Pattern matching, after all, is one of the things that frustrates me most about these AI models. It’s the engine behind the hallucinations, the confident wrong answers, the moments where a model completes a pattern with something plausible rather than something true and presents it like settled fact. It’s also, to be fair, the thing behind the responses that feel almost unreasonably good – the leaps that land precisely because the model connected dots you hadn’t seen yet. I’ve complained about the misses more times than I can count. And here was the AI telling me I was doing the very thing that frustrates me most about it.

But standing there on a dirt path watching Gracie investigate that same patch of grass for the third time, I had to sit with an uncomfortable recognition: I was doing the exact same thing. Drawing on prior conversations, previous outcomes, familiar structures. Connecting incomplete dots and filling gaps with probability-weighted guesses. The behavior I keep criticizing in machines was running quietly underneath my own thinking the entire time. That irony was not lost on me.

Here’s the thing, though – once you see it, you can’t unsee it. And I don’t just mean in that one moment on a walk. I mean everywhere. We assume tone from phrasing in an email and act on that assumption before we’ve verified it. We predict how someone will behave based on one previous interaction. We interpret silence as intent. We build entire strategic narratives from partial information and then defend them in meetings like we arrived at them through rigorous analysis. Most of the time we call this intuition or gut instinct. Sometimes, if we’re being honest, it’s just a really confident guess.

So why does it bother us so much when AI does the same thing?

I think it’s because the seams are showing. When a model makes a leap that feels slightly off, you can see it reaching – extending a pattern past where the data actually supports it. When we make the same kind of leap, it feels organic. Earned. Our patterning happens internally, wrapped in years of experience and a healthy layer of confidence. The model’s happens right out in the open where anyone can poke at it.

And honestly, that visibility is more useful than most people give it credit for.

I spend a lot of my working hours inside these tools. Building with them, stress-testing them, occasionally arguing with them. I’m building a platform called Artwell that’s rooted in the belief that AI should sharpen human storytelling rather than flatten it, so this isn’t theoretical for me – it’s the work. And the thing I didn’t expect – the thing nobody really talks about at AI conferences – is how much that daily interaction has changed the way I observe my own reasoning. Not in some grand philosophical way. In small, practical, slightly annoying ways. A model overextends on an inference, and I catch myself thinking, “Wait, didn’t I do almost exactly that in a pitch last Tuesday?” It fills in missing context with a plausible guess, and I realize I’ve been doing the same thing in emails all week without a second thought.

It’s like having a sparring partner who occasionally throws a sloppy punch and forces you to notice your own footwork isn’t as clean as you thought.

Used passively – just accepting whatever the model generates and moving on – AI absolutely reinforces lazy thinking. I’ve seen it happen. I’ve done it myself on a Thursday afternoon when I just need a first draft out the door. But used with even a little bit of intentionality, something different happens. You start noticing the gap between what you actually observed and what you concluded. You start asking yourself questions you’d normally skip: Is this conclusion real, or am I overfitting to one example? Am I projecting intent where there might just be silence? Am I mistaking familiarity for truth?

Those questions aren’t comfortable. But they’re the kind of questions that make you a sharper thinker over time. And not just analytically – the same awareness that helps you catch a bad inference in a strategy deck starts showing up in how you read a room, how you listen to a colleague who’s frustrated, how you check your own assumptions before reacting to an email that rubbed you wrong. It shows up at the dinner table when your kid says something that irritates you and you catch yourself mid-reaction, realizing you’re responding to a pattern from last week’s argument, not to what they’re actually saying right now. When you get in the habit of interrogating your own pattern matching, it doesn’t stay contained to work. It changes how you show up for the people around you.

Maybe that shouldn’t surprise us. Humans built these systems, after all. We modeled them on how we process language and probability and context. But somewhere along the way, most of us forgot that the resemblance runs both directions. AI wasn’t designed to think like a strange alien intelligence. It was designed to think like us – and the reflection it sends back, if you’re willing to look at it honestly, is more revealing than most of us expected.

The broader AI conversation right now is almost entirely about capability – what models will generate next, how fast they’ll scale, which jobs they’ll absorb. And look, I get it. That’s where the money is, and it’s where the anxiety is. But there’s a quieter shift happening underneath that nobody’s writing breathless LinkedIn posts about. As these systems become embedded in how we brainstorm, plan, and make decisions, they’re externalizing something humans have always kept hidden: the mechanics of how we actually think. They’re making it harder to pretend that what we call logic isn’t mostly structured probability. That what we call certainty isn’t compressed experience. That what we call judgment isn’t, at its core, sophisticated guesswork refined over time.

Gracie doesn’t question why she keeps returning to that same patch of grass. The pattern works often enough. She trusts it. Honestly, most days, humans aren’t that different – we’ve just built more elaborate narratives around our version of sniffing the same spot.

AI isn’t replacing thinking. It’s showing us what thinking actually looks like under the hood. And for those of us willing to sit with that instead of flinching from it, the real gift isn’t efficiency or automation. It’s the chance to become more deliberate, more honest, and maybe a little more human in how we use the minds we’ve had all along.

The next time you catch yourself making a confident assumption – in a meeting, in an email, in a conversation with someone you love – pause for a second. Ask yourself where the pattern came from. You might be surprised how often the answer is worth examining.

Tomorrow morning, I think I’ll take Gracie down a different trail. Same nose, same instincts – just new ground. She’ll have to trust the patterns she already knows in a place they’ve never been tested. Come to think of it, that makes two of us.

Share

Subscribe now

Artwell

Merry Christmas from Santa’s Vibeshop!

Share

Subscribe now

Artwell

The recent coverage of Copilot’s slow adoption has focused on familiar explanations: pricing resistance, unclear value, enterprise caution, change management. Those factors are real, but they are surface symptoms. They don’t explain why Microsoft’s AI efforts feel less convincing than they should, given the company’s scale, talent, and access to some of the most advanced models in the world.

The deeper issue sits lower in the stack. Microsoft has invested heavily in intelligence, but far less successfully in the conditions that allow intelligence to become indispensable. What users are reacting to is not a lack of capability, but a lack of cohesion. AI appears in Microsoft’s products as something added, rather than something assumed. It shows up as a panel, a button, a suggestion, a helper, not as the organizing principle of the experience itself.

This is where the idea of Artificial Integration stops being a clever phrase and starts being a diagnosis.

When Work Had a Starting Point

For most of Microsoft’s history, its strength came from owning the starting point of work. You opened Word because that was where documents began. You opened Excel because that was where analysis lived. Windows and Office were not just tools, they were defaults. Productivity flowed inward toward the platform.

Generative AI has reversed that gravity. Today, many people begin with intelligence first and worry about format later. They open ChatGPT or Claude to think, draft, summarize, or structure ideas before deciding whether a document even needs to exist. When AI sits outside the application, the application loses its privilege. Integration becomes the difference between relevance and bypass.

For most of the software era, tools defined how work happened. Interfaces imposed structure, and users adapted. That model held for decades. What AI has changed is the direction of that influence. Workflows now shape tools, not the other way around, and conversational interfaces collapse that feedback loop almost instantly. When people can express intent directly, they stop tolerating systems that require translation, navigation, or indirection.

The Sidebar Problem

This is where Microsoft’s challenge becomes visible. Copilot often feels adjacent to the work rather than embedded within it. The experience suggests augmentation instead of authorship, assistance instead of orchestration. That may have been a reasonable compromise under intense time pressure, but it carries consequences. Once users experience AI as something separate from the core workflow, they start asking a simple question: why am I here at all?

What users increasingly expect is not an AI they consult, but an application that behaves intelligently by default. Conversation has become the new entry point for work. When intelligence lives somewhere else, in a panel, a sidebar, or a separate mental step, users feel the friction immediately. They are being asked to meet the system where it is, rather than the system meeting them where they already work.

This isn’t an argument that platforms should blindly follow users. It’s an acknowledgment that the interface for intelligence has shifted. Users still work inside applications, but they expect those applications to respond directly to intent – not to route intelligence through a separate step.

The Coherence Gap

The difficulty of unwinding these decisions is amplified by ecosystem realities. Apple and Google both benefit from something Microsoft no longer has: a cohesive, end-to-end environment where hardware, operating system, and intelligence evolve together. Apple’s decision to delay public AI commitments while reinforcing its hardware ecosystem now looks prescient. When Gemini enters Apple Intelligence, it will arrive in a system already designed to feel unified. The intelligence will not compete with the interface; it will inhabit it.

Google enjoys a similar advantage through Android and Pixel. Intelligence moves with the device, not alongside it.

Microsoft, by contrast, is integrating AI into a product surface that has become fragmented across devices, platforms, and contexts. Windows no longer functions as an operating-system moat. Office no longer defines the default starting point for knowledge work. Productivity has become more fluid, more conversational, and less bound to a specific application.

In that environment, AI that feels bolted on is easy to bypass.

Smart Enough, But Not Native Enough

This is why Copilot’s adoption challenges should not be interpreted as a referendum on AI’s usefulness. They are a signal that integration quality now matters as much as intelligence itself. Users do not reject AI because it is unfamiliar. They reject it when it feels optional, interruptive, or redundant with tools they already trust.

Microsoft still has formidable strengths. Azure remains one of the most significant infrastructure moats in the industry, and the company’s enterprise footprint is unmatched. But infrastructure alone does not create habit. Habit forms where work begins. And increasingly, work begins in conversation, not in documents.

Until AI becomes the place where work starts inside Microsoft’s ecosystem, not the place users visit after the fact, adoption will continue to lag behind capability.

Beyond the Bolt-On

Artificial Integration is not a branding problem or a messaging problem. It is the structural gap between intelligence and experience, and until that gap closes, capability alone will not drive adoption.

The companies that lead the next decade of AI will not be the ones with the largest models or the loudest launches. They will be the ones that design systems where intelligence feels native, inevitable, and inseparable from the work itself. Microsoft is still capable of being one of those companies. But that future depends on whether AI remains something it adds, or something it builds around.

Bruce Bracken is the founder of Artwell.ai and former Head of Podcasts & Digital Experiences at Microsoft, where he built a content ecosystem that generated $59M in attributed Azure revenue. He writes about the gap between what AI companies say and what humans actually experience.

Share

Subscribe now

Artwell