Florian

[Florian @floismus]

@alexinexxx Using control over an AI tool to get control over more engineers :-D

[Florian @floismus]

@alecwilcock Chimera

[Florian @floismus]

@trq212 Feels rather like I work more than ever, everything is possible now, all those crazy ideas that have been just a dream two years ago

[Florian @floismus]

@thsottiaux <3 We will call this round of quota tibos not tokens :-)

[Florian @floismus]

@bcherny Turn every kind error you find into a guardrail or skill and let claude do TDD and make sure that kind of error is fixed for all of the application.
No matter if it's a architecture violation or a CSS missuse.
With that over time the vibe will produce more clean code.

[Florian @floismus]

@trq212 Love it, just had an issue where it was not my stop hook but a wrong description of the stop hook API. And Claude just suggested to send that feedback by itself, awesome! Would have liked to know what that env but no transcript data would be and if that feedback works without it.

[Florian @floismus]

@bcherny Yeah the newest thing in agentic coding is based on quite old best practices and those things we already did and loved. And agentic coding makes it so much easier to get all those guardrails in place 👍😎🥳

[Boris Cherny @bcherny]

Something I have been thinking about: in the past, the best engineers I knew spent a lot of time automating their work in various ways. Better vim/emacs automations, writing lint rules to catch repeat code issues, building up a suite of e2e tests so they don't need to smoke test the app manually. These kinds of things were the highest leverage activities an engineer could do, because it multiplied their own output, which in turn meant they could build more things.

I think many of these automations have become even more important now. This is true for a number of reasons.

First, infra and DevX automation speeds you up. And if you are running an army of agents, each of those agents will be sped up also. More automation == more output per unit of time.

Second, moving things to code improves efficiency. Your agent could fix an issue every time it sees that issue happen, but that uses tokens and might miss cases. If Claude instead writes a lint rule, CI step, or routine, that class of issue can be fully automated forever. This is really what people are talking about when they talk about loops -- it's about automating entire types of busywork rather than solving them one off. This isn't a new idea at all. Engineers have been doing this for a long time!

Third and most importantly, automation makes it possible for others to contribute to the codebase more easily. Increasingly what I am seeing is engineers are contributing to codebases on day one because Claude can navigate the codebase for them, and that non-engineers are able to contribute to a codebase as effectively as engineers can. What gets in the way of both of these is domain knowledge that lives in peoples' heads rather than in automation -- the stuff you used to have to learn when ramping up. What has changed thanks to agents is the domain knowledge that can be encoded as infrastructure is no longer limited to what is expressible in lint rules and types and tests; it can now capture nearly all domain knowledge, encoded as code comments and skills and CLAUDE.md rules and memories. If I put up a PR for an iOS codebase I don't know and a code reviewer rejects it because it doesn't use the right framework, or if a designer builds a new feature and it gets rejected because it doesn't follow the right architectural patterns, these are failures of automation.

Every team should be writing the CLAUDE.md's, REVIEW.md's, skills, and docs that enable agents to productively work in their codebase with zero additional context from the prompter. This sounds crazy, and at the same time is a natural extension of the stuff engineers have always done: automate, and encode domain knowledge as infrastructure. As the model gets smarter and as the harness matures, this task becomes easier. In the meantime, it is on every team to look for ways to convert their domain knowledge to infra so that Claude can write code better, so that code review catches issues automatically, and so the next person working on your codebase can contribute more easily.

[Florian @floismus]

Yeah! No more sleep tonight!

[ClaudeDevs @ClaudeDevs]

We've reset 5-hour and weekly rate limits for all users.

[Florian @floismus]

@_catwu Love it! When do we get Claude world, as mmo coding agent that can help to improve open source by just some messages?

[Claude @claudeai]

We're extending access to Claude Fable 5 on all paid plans through July 12.

[Florian @floismus]

@levelsio @voidfreud What about MCP servers configured in the web version - they are also available in your local installed cli - take a look at the output of `/mcp`

[Florian @floismus]

Fable is back <3

[Claude @claudeai]

Fable 5 is back.

[Florian @floismus]

Love it 💕 finally 🥳🥰🤩

[ClaudeDevs @ClaudeDevs]

Claude Desktop is now available on Linux (Ubuntu and Debian) in beta.

Alongside the browser and terminal, you now get a first-class desktop experience with Claude Code, Claude Cowork, and chat on all paid plans.

[ClaudeDevs @ClaudeDevs]

Claude Desktop is now available on Linux (Ubuntu and Debian) in beta.

Alongside the browser and terminal, you now get a first-class desktop experience with Claude Code, Claude Cowork, and chat on all paid plans.

[Florian @floismus]

Agents finally always in the background by default, love it!

I always tried to nudge my agents to use subagents for almost everything and have the main agent just coordinate those, with varying success, looking forward to seeing it integrated in the harness.

[Boris Cherny @bcherny]

In the next version of Claude Code: subagents run in the background by default, so you can keep talking to Claude while your subagents work

If you want your agent to run in the foreground, just tell Claude

[Florian @floismus]

@AnthropicAI @trq212 The moment to see that fast forward isn't the only option. US - Anthropic should seize it: one adaptable framework for recursive aligned self-improvement others can draw from.

[Florian @floismus]

@_vmlops Strong prompt-injection defense. What about extending it: a P-LLM/Q-LLM split CaMeL-style with honeypot tools? The quarantined model can't act, and a reach for a fake exfil tool is a tripwire, not a 5% classifier gap. Anyone else running this in prod?

[Florian @floismus]

Built honeypot agent tools into the Q-LLM half of a P-LLM/Q-LLM split, CaMeL style. Felt like the right move.
Is this in wide use yet?