korrents

← Topics

What do coding agents change about software work?

Top voices on coding agents: David Heinemeier Hansson, Simon Willison, Tobias Lütke, Boris Cherny, Thomas Ptacek and 4 others.

All agree the tools got good; they split on what that changes and what a person is still for.

20 dated claims from 9 people · 4 views · 2 changes of mind · most recent 26 Aug 2026

The views

Our reading of what a group of claims have in common, one level above a korrent: a korrent is our reading of what somebody said, a view is our reading of what several korrents share. The claims are printed under it, so the sentence can be checked against them.

Nobody is filed under a view — claims are. A person is named here because a dated claim of theirs sits under it, and somebody whose claims sit under two views appears under both.

  1. The gain is large, and it goes to whoever works with the agent directly

    David Heinemeier HanssonThomas PtacekBoris Cherny

  2. Writing the code was never the binding constraint, so making it cheap changes less than it looks

    David Heinemeier HanssonSunil Pai

  3. The person who checks the work is a safety property, and removing them is the risk

    Mario ZechnerSimon Willison

  4. What an agent should be given to work with is still being argued out

    David Heinemeier HanssonTobias LütkeSimon WillisonPeter Steinberger

A viewour reading

The gain is large, and it goes to whoever works with the agent directly

5 claims filed here, newest first — each in the words of whoever said it.

The very large productivity gains from agents require working with them directly; putting another human in between is too slow to preserve the gain.

Mental modelBottlenecks

  1. David Heinemeier Hansson Creator of Ruby on Rails, CTO of 37signals The revelation I've had working on Omarchy the last three months is that to get that magical 10X, 100X, in a few rare cases, 1000X productivity boost, you have to interact with the agents directly, and you cannot intermediate that bandwidth with another human because it's simply too slow. DHH: Future of Programming, AI, Agentic Engineering, Vibe Coding & Linux | Lex Fridman Podcast #501youtube.com · 26 Aug 2026 · 0:19:14 into the videoAll korrents from this video
    Their wordsThe revelation I've had working on Omarchy the last three months is that to get that magical 10X, 100X, in a few rare cases, 1000X productivity boost, you have to interact with the agents directly, and you cannot intermediate that bandwidth with another human because it's simply too slow.

    Where this was said

    Watch from 0:19:14 plays here

    ↗DHH: Future of Programming, AI, Agentic Engineering, Vibe Coding & Linux | Lex Fridman Podcast #501youtube.com

    26 Aug 2026 · video · 5h 15m · spoken · official transcript

Show all 5 claims

Agent-written pull requests already outclass those of the median human contributor to an open source project.

  1. David Heinemeier Hansson Creator of Ruby on Rails, CTO of 37signals But it actually means that if you take the median programmer and their pull requests towards an average open source project, they're already getting outclassed by agents. I would rather get an agent-written pull request to one of my projects than I'd get one written by a human, and it's not just because the quality's better. It's also because I feel a lot less bad if I just reject it. DHH: Future of Programming, AI, Agentic Engineering, Vibe Coding & Linux | Lex Fridman Podcast #501youtube.com · 26 Aug 2026 · 0:31:58 into the videoAll korrents from this video
    Their wordsBut it actually means that if you take the median programmer and their pull requests towards an average open source project, they're already getting outclassed by agents. I would rather get an agent-written pull request to one of my projects than I'd get one written by a human, and it's not just because the quality's better. It's also because I feel a lot less bad if I just reject it.

    Where this was said

    Watch from 0:31:58 plays here

    ↗DHH: Future of Programming, AI, Agentic Engineering, Vibe Coding & Linux | Lex Fridman Podcast #501youtube.com

    26 Aug 2026 · video · 5h 15m · spoken · official transcript

There is no accumulated advantage in agentic coding: someone who ignored the whole of the past year could catch up to the frontier in two weeks.

Filed here becauseAbout who the gain goes to: on this claim, anyone who starts now.

  1. David Heinemeier Hansson Creator of Ruby on Rails, CTO of 37signals There's not any accumulation, which is in some ways a great relief. DHH: Future of Programming, AI, Agentic Engineering, Vibe Coding & Linux | Lex Fridman Podcast #501youtube.com · 26 Aug 2026 · 1:24:08 into the videoAll korrents from this video
    Their wordsThere's not any accumulation, which is in some ways a great relief.

    Where this was said

    Watch from 1:24:08 plays here

    ↗DHH: Future of Programming, AI, Agentic Engineering, Vibe Coding & Linux | Lex Fridman Podcast #501youtube.com

    26 Aug 2026 · video · 5h 15m · spoken · official transcript

Coding agents make every plugin interface you ever blew off trivially usable, which is a large productivity win.

  1. Thomas Ptacek Security engineer; founder of Matasano Security and Latacora I can't say enough how excellent a coding agent is at customizing Chrome with extensions. A huge productivity win. Every plugin interface you've ever blown off is now trivially usable. @tqbf on Xx.com · 19 Aug 2026
    Their wordsI can't say enough how excellent a coding agent is at customizing Chrome with extensions. A huge productivity win. Every plugin interface you've ever blown off is now trivially usable.

    ↗@tqbf on Xx.com

    19 Aug 2026

With a good enough plan, a coding agent implements the whole change in one attempt.

  1. Boris Cherny Creator of Claude Code at Anthropic once there is a good plan, it will one-shot the implementation almost every time. Building Claude Code with Boris Cherny — The Pragmatic Engineernewsletter.pragmaticengineer.com · 4 Mar 2026
    Their wordsonce there is a good plan, it will one-shot the implementation almost every time.

    Where this was said

    ↗Building Claude Code with Boris Cherny — The Pragmatic Engineernewsletter.pragmaticengineer.com

    4 Mar 2026 · spoken · official transcript

A viewour reading

Writing the code was never the binding constraint, so making it cheap changes less than it looks

4 claims filed here, newest first — each in the words of whoever said it.

Most organisations are bottlenecked on ideas, vision and taste rather than on implementation, so making code cheap does not make them productive.

Mental modelBottlenecks

  1. David Heinemeier Hansson Creator of Ruby on Rails, CTO of 37signals The other thing I'd say is that most organizations don't know what they want. They don't know how to make it better. They're not bottlenecked on implementation. They're bottlenecked on ideas. They're bottlenecked on vision. They're bottlenecked on taste. And if you don't have those element in excess of your implementational capacity, it doesn't help. So you can make a lot of shitty ideas come true, then what? DHH: Future of Programming, AI, Agentic Engineering, Vibe Coding & Linux | Lex Fridman Podcast #501youtube.com · 26 Aug 2026 · 0:20:13 into the videoAll korrents from this video
    Their wordsThe other thing I'd say is that most organizations don't know what they want. They don't know how to make it better. They're not bottlenecked on implementation. They're bottlenecked on ideas. They're bottlenecked on vision. They're bottlenecked on taste. And if you don't have those element in excess of your implementational capacity, it doesn't help. So you can make a lot of shitty ideas come true, then what?

    Where this was said

    Watch from 0:20:13 plays here

    ↗DHH: Future of Programming, AI, Agentic Engineering, Vibe Coding & Linux | Lex Fridman Podcast #501youtube.com

    26 Aug 2026 · video · 5h 15m · spoken · official transcript

Show all 4 claims

The ideas in software originate with humans; agents build what people tell them to build.

Filed here becauseFiled here as the strong form of the same argument — and its holder has since stepped back from it.

  1. David Heinemeier Hansson Creator of Ruby on Rails, CTO of 37signals I thought all the ideas would originate with the humans. They would tell their agents what to build, and off they went. DHH: Future of Programming, AI, Agentic Engineering, Vibe Coding & Linux | Lex Fridman Podcast #501youtube.com · 26 Aug 2026 · 0:36:24 into the videoAll korrents from this video

    Recalls holdingI thought all the ideas would originate with the humans. They would tell their agents what to build, and off they went.

    Their words nowI don't think that's true anymore at all.

    Where this was said

    Watch from 0:36:24 plays here

    ↗DHH: Future of Programming, AI, Agentic Engineering, Vibe Coding & Linux | Lex Fridman Podcast #501youtube.com

    26 Aug 2026 · video · 5h 15m · spoken · official transcript

Coding agents made implementation cheap without making roadmaps shorter or software better -- the binding constraint was never how fast code gets written.

Mental modelBottlenecks

  1. Sunil Pai Systems and developer-tools engineer at Cloudflare as of 2026 buuuut my roadmap is’nt ten times shorter. and I’m definitely not ten times better at deciding what is worth building. teams haven’t started casually shipping a year of product work every month. you can look around and and it’s hard to say what software has gotten meaningfully better in the last year. the task isn't the jobsunilpai.dev · 14 Aug 2026
    Their wordsbuuuut my roadmap is’nt ten times shorter. and I’m definitely not ten times better at deciding what is worth building. teams haven’t started casually shipping a year of product work every month. you can look around and and it’s hard to say what software has gotten meaningfully better in the last year.

    ↗the task isn't the jobsunilpai.dev

    14 Aug 2026

A machine can complete the task and completely miss the job -- coding makes this easy to see precisely because its tasks are so legible and verifiable.

  1. Sunil Pai Systems and developer-tools engineer at Cloudflare as of 2026 a cybernetic machine can nail the task and completely miss the job. coding makes this very very easy to see because the tasks are so legible (and verifiable!): write the function, migrate the API, fix the tests. the task isn't the jobsunilpai.dev · 14 Aug 2026
    Their wordsa cybernetic machine can nail the task and completely miss the job. coding makes this very very easy to see because the tasks are so legible (and verifiable!): write the function, migrate the API, fix the tests.

    ↗the task isn't the jobsunilpai.dev

    14 Aug 2026

A viewour reading

The person who checks the work is a safety property, and removing them is the risk

4 claims filed here, newest first — each in the words of whoever said it.

A coding agent does not learn from its mistakes the way a person does -- it repeats the same error indefinitely unless a human notices and writes it down.

  1. Mario Zechner Austrian developer who created the libGDX game framework An agent has no such learning ability. At least not out of the box. It will continue making the same errors over and over again. Depending on the training data it might also come up with glorious new interpolations of different errors. Thoughts on slowing the fuck downmariozechner.at · 25 Mar 2026
    Their wordsAn agent has no such learning ability. At least not out of the box. It will continue making the same errors over and over again. Depending on the training data it might also come up with glorious new interpolations of different errors.

    ↗Thoughts on slowing the fuck downmariozechner.at

    25 Mar 2026

Show all 4 claims

The human bottleneck is a safety property: remove it and small, individually harmless coding errors compound at a rate nobody feels until it is too late.

Mental modelCompounding

  1. Mario Zechner Austrian developer who created the libGDX game framework With an orchestrated army of agents, there is no bottleneck, no human pain. These tiny little harmless booboos suddenly compound at a rate that's unsustainable. You have removed yourself from the loop, so you don't even know that all the innocent booboos have formed a monster of a codebase. You only feel the pain when it's too late. Thoughts on slowing the fuck downmariozechner.at · 25 Mar 2026
    Their wordsWith an orchestrated army of agents, there is no bottleneck, no human pain. These tiny little harmless booboos suddenly compound at a rate that's unsustainable. You have removed yourself from the loop, so you don't even know that all the innocent booboos have formed a monster of a codebase. You only feel the pain when it's too late.

    ↗Thoughts on slowing the fuck downmariozechner.at

    25 Mar 2026

In an agent-written codebase the agent-written tests are equally untrustworthy; manually using the product is the only reliable measure of whether it works.

  1. Mario Zechner Austrian developer who created the libGDX game framework Worse, you realize that the gazillions of unit, snapshot, and e2e tests you had your clankers write are equally untrustworthy. The only thing that's still a reliable measure of "does this work" is manually testing the product. Thoughts on slowing the fuck downmariozechner.at · 25 Mar 2026
    Their wordsWorse, you realize that the gazillions of unit, snapshot, and e2e tests you had your clankers write are equally untrustworthy. The only thing that's still a reliable measure of "does this work" is manually testing the product.

    ↗Thoughts on slowing the fuck downmariozechner.at

    25 Mar 2026

Running coding agents without approval prompts is a normalization-of-deviance risk: the fact that it has not caused a disaster yet is itself the problem.

Mental modelMargin of safety

  1. Simon Willison Co-creator of Django and creator of Datasette I run in YOLO mode all the time, despite being deeply aware of the risks involved. It hasn’t burned me yet... ... and that’s the problem. 2025: The year in LLMssimonwillison.net · 31 Dec 2025
    Their wordsI run in YOLO mode all the time, despite being deeply aware of the risks involved. It hasn’t burned me yet... ... and that’s the problem.

    ↗2025: The year in LLMssimonwillison.net

    31 Dec 2025

A viewour reading

What an agent should be given to work with is still being argued out

5 claims filed here, newest first — each in the words of whoever said it.

Overly prescriptive instructions make a coding agent worse, for the same reason a micromanaging boss makes a programmer worse.

  1. David Heinemeier Hansson Creator of Ruby on Rails, CTO of 37signals it was actually being damaged by overly prescriptive humans. And any programmer who's had a pointy-haired boss knows exactly what that is like. When the boss walks into the room, doesn't know anything, starts telling you how to program, how to code. What do you do? You sulk. You write shittier code if you're mandated to do things that are against your better judgment. Why would an agent not be the same? DHH: Future of Programming, AI, Agentic Engineering, Vibe Coding & Linux | Lex Fridman Podcast #501youtube.com · 26 Aug 2026 · 0:56:38 into the videoAll korrents from this video
    Their wordsit was actually being damaged by overly prescriptive humans. And any programmer who's had a pointy-haired boss knows exactly what that is like. When the boss walks into the room, doesn't know anything, starts telling you how to program, how to code. What do you do? You sulk. You write shittier code if you're mandated to do things that are against your better judgment. Why would an agent not be the same?

    Where this was said

    Watch from 0:56:38 plays here

    ↗DHH: Future of Programming, AI, Agentic Engineering, Vibe Coding & Linux | Lex Fridman Podcast #501youtube.com

    26 Aug 2026 · video · 5h 15m · spoken · official transcript

Show all 5 claims

Coding agents should read the shared AGENTS.md convention; insisting on a tool-specific CLAUDE.md creates split-brain problems across a team.

  1. Tobias Lütke Co-founder and CEO of Shopify I’m thinking about banning Claude code at Shopify until they change their mind and read AGENTS.md and .agents/skills etc. Insisting on only reading CLAUDE.md sometimes leads to split brain problems when different team members use different tools. Just unnecessary. @tobi on Xx.com · 25 Aug 2026
    Their wordsI’m thinking about banning Claude code at Shopify until they change their mind and read AGENTS.md and .agents/skills etc. Insisting on only reading CLAUDE.md sometimes leads to split brain problems when different team members use different tools. Just unnecessary.

    ↗@tobi on Xx.com

    25 Aug 2026

Claude Code is the coding agent to build with.

  1. Pieter Levels · held 8 weeks Dutch bootstrapper who builds and runs his internet startups alone I won't use SnekAI but I'll probably switch to @xAI soon Claude Code treats you like a child getting lootboxes so I'll probably switch to xAIlevels.io · 24 Aug 2026

    No longer holdsI won't use SnekAI but I'll probably switch to @xAI soon

    Their words nowYou're treated like a child getting lootboxes just to be able to do your work while you pay massive amounts of money for extra usage tokens

    ↗Claude Code treats you like a child getting lootboxes so I'll probably switch to xAIlevels.io

    24 Aug 2026

MCP may turn out to be a one-year wonder, because coding agents that can run arbitrary shell commands make Bash the best general-purpose tool interface.

  1. Simon Willison Co-creator of Django and creator of Datasette The reason I think MCP may be a one-year wonder is the stratospheric growth of coding agents. It appears that the best possible tool for any situation is Bash—if your agent can run arbitrary shell commands, it can do anything that can be done by typing commands into a terminal. 2025: The year in LLMssimonwillison.net · 31 Dec 2025
    Their wordsThe reason I think MCP may be a one-year wonder is the stratospheric growth of coding agents. It appears that the best possible tool for any situation is Bash—if your agent can run arbitrary shell commands, it can do anything that can be done by typing commands into a terminal.

    ↗2025: The year in LLMssimonwillison.net

    31 Dec 2025

MCP servers are worth removing outright: an agent reading the code directly is faster, and pollutes its context less, than one that can reach for a tool unasked.

  1. Peter Steinberger Founded PSPDFKit in 2011 and ran it for a decade Even removed my last MCP, since Claude sometimes would go off spinning up Playwright unasked when it could simply read the code - which is faster and pollutes the context less. My Current AI Dev Workflowsteipete.me · 25 Aug 2025
    Their wordsEven removed my last MCP, since Claude sometimes would go off spinning up Playwright unasked when it could simply read the code - which is faster and pollutes the context less.

    ↗My Current AI Dev Workflowsteipete.me

    25 Aug 2025

Read, and under none of the views

On the topic, but saying something the views above do not cover.

Agentic code search has recall that falls as a codebase grows, so the agent does not find the code it needs -- and which search tool you give it barely matters.

  1. Mario Zechner Austrian developer who created the libGDX game framework You can give it a Bash tool so it can ripgrep its way through the codebase. You can give it some queryable codebase index, an LSP server, a vector database. In the end it doesn't matter much. The bigger the codebase, the lower the recall. Low recall means that your agent will, in fact, not find all the code it needs to do a good job. Thoughts on slowing the fuck downmariozechner.at · 25 Mar 2026
    Their wordsYou can give it a Bash tool so it can ripgrep its way through the codebase. You can give it some queryable codebase index, an LSP server, a vector database. In the end it doesn't matter much. The bigger the codebase, the lower the recall. Low recall means that your agent will, in fact, not find all the code it needs to do a good job.

    ↗Thoughts on slowing the fuck downmariozechner.at

    25 Mar 2026

The part of a coding agent that actually acts on code is not AI at all — it is simple systems code wired to ground truth, the way a Makefile is.

  1. Thomas Ptacek Security engineer; founder of Matasano Security and Latacora The code in an agent that actually “does stuff” with code is not, itself, AI. This should reassure you. It’s surprisingly simple systems code, wired to ground truth about programming in the same way a Makefile is. My AI Skeptic Friends Are All Nutsfly.io · 2 Jun 2025
    Their wordsThe code in an agent that actually “does stuff” with code is not, itself, AI. This should reassure you. It’s surprisingly simple systems code, wired to ground truth about programming in the same way a Makefile is.

    ↗My AI Skeptic Friends Are All Nutsfly.io

    2 Jun 2025

Where the claims collide

Asserted relations, each with the curator’s reason — never computed, because similarity cannot tell agreement from opposition.

A machine can complete the task and completely miss the job -- coding makes this easy to see precisely because its tasks are so legible and verifiable.

held by Sunil Pai

These pull against each other

With a good enough plan, a coding agent implements the whole change in one attempt.

held by Boris Cherny

Cherny says a good enough plan gets a coding agent to implement the whole change in one attempt. Pai says a machine can complete the task and completely miss the job, and that coding makes this easy to see precisely because its tasks are so legible. The plan is what makes the task legible, which is the same property that lets the job go missing.

A coding agent does not learn from its mistakes the way a person does -- it repeats the same error indefinitely unless a human notices and writes it down.

held by Mario Zechner

These pull against each other

Agent-written pull requests already outclass those of the median human contributor to an open source project.

held by David Heinemeier Hansson

Hansson says agent-written pull requests already outclass the median human contributor to an open source project. Zechner says an agent repeats the same error indefinitely unless a person notices and writes it down. A contributor who cannot learn from review is a different kind of contributor from the median human, however good the first patch is.

Who has changed their mind

  1. David Heinemeier Hansson ·

    No longer holdsThe ideas in software originate with humans; agents build what people tell them to build.

    On our filingfrom Writing the code was never the binding constraint, so making it cheap changes less than it looks to The gain is large, and it goes to whoever works with the agent directly

  2. Pieter Levels · · held 8 weeks

    No longer holdsClaude Code is the coding agent to build with.

On this topic, not yet placed

Matched by wording, not yet read into a view. The archive grows on its own, so this is where new claims wait for somebody to read them.

The full chronology, with every quote: coding agents as a timeline · person by person, side by side · all topics · Something wrong? Tell us