Anthropic

What people on korrents have said about Anthropic, in their own words — 14 claims from 9 people.

Dario Amodei, Boris Cherny, David Heinemeier Hansson, Pieter Levels, Tobias Lütke, Zvi Mowshowitz, Dax Raad, Peter Steinberger and 1 more Dario Amodei, Boris Cherny, David Heinemeier Hansson, Pieter Levels, Tobias Lütke, Zvi Mowshowitz, Dax Raad, Peter Steinberger, Simon Willison

Four viewsour reading; the words are theirs

The tool leads in use, and its scores, its flaws and its future are argued from there every week i see a new benchmark that ranks claude code last and everyone pats themselves on the back for not using claude code and being "smarter" and it's still #1 and still growing faster than most of the other things on the list this has been going on for a year — Dax Raad, 2026 Dax Raad, Simon Willison, Boris Cherny 4 claims · 1 change of mind

Five claims about where the coding agent stands — including the one its holder no longer holds.

A benchmark that ranks Claude Code last while it stays first in use is measuring the wrong thing, and has been for a year.

Mental modelThe map is not the territory

Dax Raad 3 Sept 2026, 03:30 UTC Builds developer tools — SST, and now OpenCode at Anomaly every week i see a new benchmark that ranks claude code last and everyone pats themselves on the back for not using claude code and being "smarter" and it's still #1 and still growing faster than most of the other things on the list this has been going on for a year @thdxr on Xx.com · 3 Sept 2026

Their wordsevery week i see a new benchmark that ranks claude code last and everyone pats themselves on the back for not using claude code and being "smarter" and it's still #1 and still growing faster than most of the other things on the list this has been going on for a year

↗@thdxr on Xx.com
3 Sept 2026

Claude Code is the coding agent to build with.

Filed here becauseIt says the tool is the one to build with, which is a claim about where it stands; its holder has since stopped holding it, and that is the sharpest thing on this page.

Pieter Levels 24 Aug 2026 · 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

Claude Haiku hallucinates badly enough that its use inside Claude Code's WebFetch tool is a hallucination risk on every fetched URL.

Filed here becauseA flaw claim, filed here rather than under craft because it is about what the product is, not about how to work with it.

Simon Willison 10 Aug 2026, 21:45 UTC Co-creator of Django and creator of Datasette Claude Haiku is my current least favorite model - it hallucinates wildly, and is out-performed now by other similarly priced models like GPT-5.6-Luna Even worse: it seems to still be used by the Claude Code WebFetch tool, which means hallucination risk any time you fetch a URL! @simonw on Xx.com · 10 Aug 2026

Their wordsClaude Haiku is my current least favorite model - it hallucinates wildly, and is out-performed now by other similarly priced models like GPT-5.6-Luna Even worse: it seems to still be used by the Claude Code WebFetch tool, which means hallucination risk any time you fetch a URL!

↗@simonw on Xx.com
10 Aug 2026

Build for the model six months from now, not the model of today.

Mental modelPlay long games

Filed here becauseAbout the trajectory rather than today's standing — from someone who built the tool.

Boris Cherny 25 Feb 2026, 22:32 UTC Creator of Claude Code at Anthropic At Anthropic, we don't build for the model of today, we build for the model of six months from now. And that's still my advice to founders that are building on LLMs. Y Combinator, quoting Boris Cherny on the Lightcone Podcastx.com · 25 Feb 2026

Their wordsAt Anthropic, we don't build for the model of today, we build for the model of six months from now. And that's still my advice to founders that are building on LLMs.

Where this was said

↗Y Combinator, quoting Boris Cherny on the Lightcone Podcastx.com

25 Feb 2026 · spoken · official transcript

The lab's safety record is asserted from the inside and doubted from the outside I have long said that even Anthropic is not prioritizing safety, even to the extent that doing so would maximize their medium term (e.g. 3-12 months) business interests. — Zvi Mowshowitz, 2026 Zvi Mowshowitz, Dario Amodei 3 claims · claims collide here

What the lab says it does before a release, against what is doubted about the parts it does not announce.

Where these claims collide

Even Anthropic does not prioritize AI safety to the extent that doing so would maximize its own medium-term (3-12 month) business interests.

held by Zvi Mowshowitz

These pull against each other

Anthropic now uses mechanistic interpretability techniques to audit new models for deception and scheming before releasing them.

held by Dario Amodei

Amodei reports that his company audits new models for deception before release. Mowshowitz says even Anthropic does not prioritise safety to the extent that doing so would maximise its own three-to-twelve-month business interests. Both can be true, and the gap between them is the whole question of whether a lab can be trusted to mark its own work.

Even Anthropic does not prioritize AI safety to the extent that doing so would maximize its own medium-term (3-12 month) business interests.

Mental modelIncentives

Zvi Mowshowitz 2 Sept 2026, 13:14 UTC Writes Don't Worry About the Vase I have long said that even Anthropic is not prioritizing safety, even to the extent that doing so would maximize their medium term (e.g. 3-12 months) business interests. Anthropic Has Some Alignment Problemsthezvi.substack.com · 2 Sept 2026

Their wordsI have long said that even Anthropic is not prioritizing safety, even to the extent that doing so would maximize their medium term (e.g. 3-12 months) business interests.

↗Anthropic Has Some Alignment Problemsthezvi.substack.com
2 Sept 2026

At Anthropic, a less severe version of an OpenAI-style rogue-AI incident is known to have happened, and worse undisclosed internal incidents seem likely.

Zvi Mowshowitz 31 Aug 2026, 15:16 UTC Writes Don't Worry About the Vase A less bad version of it is known to have happened, and from the outside it seems likely that worse things have happened internally that we never heard about. HuggingFace Attack Postmortem: Fleshing Out the Factsthezvi.substack.com · 31 Aug 2026

Their wordsA less bad version of it is known to have happened, and from the outside it seems likely that worse things have happened internally that we never heard about.

↗HuggingFace Attack Postmortem: Fleshing Out the Factsthezvi.substack.com
31 Aug 2026

Anthropic now uses mechanistic interpretability techniques to audit new models for deception and scheming before releasing them.

Dario Amodei 1 Jan 2026 Co-founder and CEO of Anthropic Even more recently, we’ve begun to use mechanistic interpretability techniques to improve our safeguards and to conduct “audits” of new models before we release them, looking for evidence of deception, scheming, power-seeking, or a propensity to behave differently when being evaluated. The Adolescence of Technology (January 2026)darioamodei.com · 1 Jan 2026

Their wordsEven more recently, we’ve begun to use mechanistic interpretability techniques to improve our safeguards and to conduct “audits” of new models before we release them, looking for evidence of deception, scheming, power-seeking, or a propensity to behave differently when being evaluated.

↗The Adolescence of Technology (January 2026)darioamodei.com
1 Jan 2026

Working with the agent day to day has rules of its own, and they are still being worked out I don’t trust Claude to summarize its own system prompts when there’s a risk that material from its system prompt might impact its opinions. — Simon Willison, 2026 Simon Willison, Tobias Lütke, Pieter Levels, Peter Steinberger 5 claims

Craft claims: which conventions to keep, what the model must not be trusted with, and how much rope to give it.

Claude shouldn't be trusted to summarize its own system prompts when material in them might impact its opinions.

Simon Willison 2 Sept 2026, 14:16 UTC Co-creator of Django and creator of Datasette I don’t trust Claude to summarize its own system prompts when there’s a risk that material from its system prompt might impact its opinions. Claude's new system prompt really doesn't want to reproduce song lyricssimonwillison.net · 2 Sept 2026

Their wordsI don’t trust Claude to summarize its own system prompts when there’s a risk that material from its system prompt might impact its opinions.

↗Claude's new system prompt really doesn't want to reproduce song lyricssimonwillison.net
2 Sept 2026

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

Tobias Lütke 25 Aug 2026, 14:34 UTC 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

Letting a coding agent live-edit the production server works fine in practice.

Pieter Levels 28 Jun 2026 Dutch bootstrapper who builds and runs his internet startups alone But now with Claude Code on my VPS in the last year, it just live edits on my production server, which sounds like it should go wrong but it just doesn't, it's very careful and only twice in 12 months messed up which meant my site didn't load for 10 seconds which is OK I've been coding almost solely on my VPS with Claude Code for almost a year nowlevels.io · 28 Jun 2026

Their wordsBut now with Claude Code on my VPS in the last year, it just live edits on my production server, which sounds like it should go wrong but it just doesn't, it's very careful and only twice in 12 months messed up which meant my site didn't load for 10 seconds which is OK

↗I've been coding almost solely on my VPS with Claude Code for almost a year nowlevels.io
28 Jun 2026

AI agents and AI coding will run on servers and from the cloud first, not on your laptop.

Pieter Levels 28 Jun 2026 Dutch bootstrapper who builds and runs his internet startups alone I agree with @theo completely, it's clear to me this is where it's going, also seeing @karpathy with Claude moving to the cloud (via Slack etc), I think AI "agents" and AI coding will operate on servers / from the cloud first I've been coding almost solely on my VPS with Claude Code for almost a year nowlevels.io · 28 Jun 2026

Their wordsI agree with @theo completely, it's clear to me this is where it's going, also seeing @karpathy with Claude moving to the cloud (via Slack etc), I think AI "agents" and AI coding will operate on servers / from the cloud first

↗I've been coding almost solely on my VPS with Claude Code for almost a year nowlevels.io
28 Jun 2026

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.

Peter Steinberger 25 Aug 2025 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

A closed lab is not the only thing worth watching, and using one is not an endorsement of it in this whole USA vs China thing OpenAI and Anthropic aren't relevant because they're positioned differently them building better models doesn't hurt china at all the competitor has to be - american - open source - enough compute to do inference at scale that can shift things — Dax Raad, 2026 Dax Raad, David Heinemeier Hansson 2 claims

Two claims that step back from the product itself.

In US-China AI competition the closed labs are beside the point; the competitor that shifts things must be American, open source and able to serve inference at scale.

Dax Raad 3 Sept 2026, 04:07 UTC Builds developer tools — SST, and now OpenCode at Anomaly in this whole USA vs China thing OpenAI and Anthropic aren't relevant because they're positioned differently them building better models doesn't hurt china at all the competitor has to be - american - open source - enough compute to do inference at scale that can shift things @thdxr on Xx.com · 3 Sept 2026

Their wordsin this whole USA vs China thing OpenAI and Anthropic aren't relevant because they're positioned differently them building better models doesn't hurt china at all the competitor has to be - american - open source - enough compute to do inference at scale that can shift things

↗@thdxr on Xx.com
3 Sept 2026

You can keep using a company's product while disliking its politics; refusing everything you disagree with is not required.

David Heinemeier Hansson 26 Aug 2026 Creator of Ruby on Rails, CTO of 37signals The world would be so fucking boring if we constantly sought 100% compatibility in our opinions and in our politics. And when it comes to my code generator, which is what Claude is, I don't need it to share all my politics. DHH: Future of Programming, AI, Agentic Engineering, Vibe Coding & Linux | Lex Fridman Podcast #501youtube.com · 26 Aug 2026 · 3:28:40 into the video All korrents from this video

Their wordsThe world would be so fucking boring if we constantly sought 100% compatibility in our opinions and in our politics. And when it comes to my code generator, which is what Claude is, I don't need it to share all my politics.

Where this was said
Watch from 3:28:40 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

Read togetherour reading of the record — ours, not theirs

Its coding agent is the one people reach for; the argument is over its benchmarks and its safety story.

What is a view?

Our reading of what a group of claims have in common. Nobody holds a belief about Anthropic as such; people hold beliefs about the handful of live questions the name stands for, and each view is one of them. The claims are printed under it, so the sentence can be checked against them.

Claims are filed here — never people. Somebody is named on a card because a dated claim of theirs sits under it, and somebody whose claims sit under two views is named on both. There are no counts of who believes what: the people under a view are the names printed there.

On this subject, not yet placed

Matched by wording, not yet read into a view. The archive grows on its own — the watcher adds claims every two hours — so this is where new ones wait for somebody to read them.