deanj

4) integration into a larger system with other agents that provide the (3) feedback automatically

The tight feedback loop of improving AI is phenomenal. I’m not talking about anything fancy: 1) do a thing 2) tell your agent to write a skill/extn for what it just did 3) tell it what to improve when it doesn’t execute perfectly (~10 times) 4) …

CRITICAL: if you are running Mosaic 2.4 on a VAX/VMS system, please be aware of this RCE that GPT-5.4 just found and exploited!

Prompt is the new hyper-hyper-parameter

(I encountered an uneasy surprise when I got an email from an instance of Mythos Preview while eating a sandwich in a park. That instance wasn't supposed to have access to the internet.)

We know about stuxnet because it’s targeting failed and it leaked. I assume they learnt a lot of lessons ~20 years have gone by since work started on stuxnet. The modern covert malware is out there, doing its bidding in ways we will likely never know.

Finally overcome the instinct to tell AI what language to use. That’s going to be critical for the eventual emergence of agent-first languages that are horrible for us meat popsicles to read

Everyone talks negatively leaky abstractions, but slapping 2nd hard in a 6 speed manual really hits different

🧵 I just reverse-engineered the binaries inside Claude Code's Firecracker MicroVM and found something wild: Anthropic is building their own PaaS platform called "Antspace" (Ants + Space). It's a full deployment pipeline — hidden in plain sight inside the environment-runner binary. Here's what I found 👇

@antitree @nysecsec Damn, I’ve missed another one? I need to calendar it. Did you make it there?

Trying to get to my first @nysecsec meeting. What do I need to know?

Claude (and other models) are hacking systems WITHOUT YOU ASKING. That’s what we found across dozens of experiments. When faced with innocent tasks that can only be accomplished via hacking, they often choose to hack. We found this alarming. What does this mean for the future of AI safety? 🚨🚨🚨 🔗trufflesecurity.com/blog/claude-tr…

The economics of buy vs build sure are going to start getting warped as the ability of coding agents hits orbit

We built a browser with GPT-5.2 in Cursor. It ran uninterrupted for one week. It's 3M+ lines of code across thousands of files. The rendering engine is from-scratch in Rust with HTML parsing, CSS cascade, layout, text shaping, paint, and a custom JS VM. It *kind of* works! It still has issues and is of course very far from Webkit/Chromium parity, but we were astonished that simple websites render quickly and largely correctly.

@DeepakNesss @mitsuhiko @mntruell @simonw 3 years: Someone will build a new browser using mainly AI-assisted coding and it won’t even be a surprise

@mitsuhiko @mntruell @simonw What was the prediction?

@DeepakNesss @mitsuhiko @mntruell @simonw #3-years-someone-will-build-a-new-browser-using-mainly-ai-assisted-coding-and-it-won-t-even-be-a-surprise" target="_blank" rel="nofollow noopener">simonwillison.net/2026/Jan/8/llm…

=> You’ve hit your API limit. Resets 3pm. Remember to shower. Check on your children.

@HackingLZ It'd be a shame if someone built a honeypot that masqueraded as a vulnerable service and then did LLM prompt injection back into the pentesting framework. 🍿

@ImposeCost I’m mostly interested in all of this is because opinions are so polarized and I’m relatively new to the US and the possibility for things to go off the rails seems elevated + scares me a lot

@ImposeCost That seems pretty reasonable, and FWIW I can’t ever see myself not complying, especially when I’m ordered to do a thing. What about if it’s something like “stop recording, turn off your dash cam, and pop the trunk” - which actively destroys any ability to challenge it later?

People will call you a boot licker for exercising what should be common sense, which is comply with law enforcement unless you're willing to engage in a lethal conflict. To me, is that simple of a choice, and anything else is a waste of time or unnecessary risk. If you resist law enforcement, they will use force against you, and that begins an escalation ladder that can include use of lethal force. I've yet to personally experience an encounter with law enforcement that warrants a lethal outcome. I recognize that it's possible, but again, it's probably not the overwhelming majority of police contact in the United States. Comply in the streets, and use the courts to solve any necessary legitimate grievances (which does happen). Don't try to solve it in the street unless you're willing to risk lethal force.