Tim Abbott

@DavidSKrueger I get plenty of initial disagreement and useful context, but all chatbots do seem to often come around to whatever my bias is with more turns. Anyway an interesting corollary is to not count someone else's AI sessions as evidence on a topic, unless you trust their judgement.

It feels like 100% of the time to me.

We're moving off X! Please follow us at linkedin.com/company/zulip-…, @zulip" target="_blank" rel="nofollow noopener">fosstodon.org/@zulip or bsky.app/profile/zulip.…. Check out those places for a major release announcement.

@zeeg Zulip may not quite qualify as big but I think it'd be a great target for this.

What’s a big public codebase that has a large api surface or otherwise server code? Preferably open source. I’d like to try running an experiment with Warden on it. If it works I’ll send you a bunch of disclosures or potential vulnerabilities.

@MicahBerkley @AnthropicAI I had no issues doing my routine "new model release" security audits of Zulip with 4.7 this week.

So @AnthropicAI is no longer allowing me to scan my own software for security vulnerabilities using Opus 4.7. This is a huge problem. Opus 4.6 this was never an issue. And respectfully Opus 4.6 was a beast at this. I'm really disappointed. Especially since I'm paying $200 month for this. I'm not going to use freaking Sonnet to do security work. @bcherny help us out maannn...

@simonw I've found out to be quite good at making changes to Zulip. I think it does matter how nice your codebase is. If you have lots of uncommented nonlocal dependencies, I'm sure you'll feel that when trying to have AI make changes, just like if you hired a new person.

Is there still a widespread belief that LLMs and coding agents are good for greenfield development but don't help for maintaining large existing codebases? I don't think that idea holds up any more

We wrote up everything we do to secure our open source projects at Astral

@simonw The thing I don't think has gotten enough attention is the upcoming catch-22 of wanting to install dependency version updates immediately because exploit generation is really fast, and supply chain attacks.

Wrote up some thoughts on Anthropic's Project Glassing, where their latest Opus-beating model is available to partnered security research organizations only Given recent alarm bells raised by credible security voices I think this is a justified decision simonwillison.net/2026/Apr/7/pro…

Proud to work at Anthropic today.

Claude is #1 in the App Store today — I want to say a huge thank you to all of our new (and existing!) users for the support. We’re working hard for you, please share your thoughts and feedback along the way.

I believe these are reasonable lines to hold. And I'm proud to work for a company willing to hold them.

Statement from Max @Tegmark, Founder and Chair of the Future of Life Institute, in the aftermath of @AnthropicAI refusing the Department of War's ultimatum: “Fully autonomous weapons systems and Orwellian AI-enabled domestic mass surveillance are affronts to our dignity and liberty. We highly commend Anthropic, OpenAI and leading researchers from across AI companies for standing up for the principle that AI should never be used to kill people without meaningful human control, and that domestic mass surveillance of US citizens is a red line that should never be crossed. We call on all AI companies to follow suit. However, our safety and basic rights must not be at the mercy of a company's internal policy; lawmakers must work to codify these overwhelmingly popular red lines into law. All AI systems should be under meaningful human control. This is especially true for those that could be used in the taking of human lives. Moreover, current AI systems are inherently unpredictable and fundamentally brittle, unsuited for very high stakes applications. Even if they could be made effective, fully autonomous weapons would pose a threat not just to human dignity and liberty but to American national security: they could inadvertently fuel escalation, and would easily proliferate, putting cheap, accessible, weapons of assassination and mass destruction in the hands of non-state actors and adversaries. They should be prohibited by the US and globally.”

Talking of spines, Anthropic definitely has one. May they continue to expose and resist the Pentagon’s blackmail.

Anthropic 2024: You can trust that we'll keep all our safety promises Antropic 2026: Nvm

"Documents that write themselves" That's the tagline for our newest feature: Town Docs. It's changed the way I write documents and emails as it supercharges the Town Assistant and makes it trivial to pull in context from all of my work to create the perfect content. Use cases from the last few days: - Pull together every piece of customer feedback from Slack, email, and support channels into a prioritized doc. After working on the doc for a bit, I was able to add things to Slack simply by saying "Create or bump tickets for every P1 and P2 item." - Automatically creating an agenda for our weekly meeting, along with references to major PRs of what we shipped last week. - Weekly review of what other AI companies are up to, which I then iterate on to figure out what we can learn from other great teams. Check it out at town.com/features/town-…!

OpenAI has dropped safety from its mission statement – can you spot another change? Old: "OpenAIs mission is to build general­ purpose artificial intelligence (AI) that safely benefits humanity, unconstrained by a need to generate financial return. [...]" New: "OpenAIs mission is to ensure that artificial general intelligence benefits all of humanity" (IRS evidence in comments)

🚨 "[@SamA] is telling me that my 3-year-old son has only two choices in life: put electrodes in his head or never get a job, become obsolete." -FLI co-founder and president @Tegmark at Florida Gov. @RonDeSantis' roundtable on AI policy earlier this week:

We just released Zulip Server 11.5; this is a security fix, along with a bunch of bugfixes. Full details in our blog post: blog.zulip.com/2026/02/05/zul…

I was just laid off by The Washington Post in the middle of a warzone. I have no words. I'm devastated.

Great 18-minute rant on why AI making art isn’t the same as a person making art, even if the audience can’t tell whether a person or an AI made it. Also applies to the conundrum of junior vs. senior workers — if AI replaces the juniors, the seniors will eventually retire, and then what? youtu.be/mb3uK-_QkOo