Extremely strong agree. This post is about adding an Edit button, but imo it is true about any possible new feature, even the most innocuous. Attack it first, if it survives, float it to the community as a possibility.

A security expert using Mythos compromised MacOS. www.wsj.com/tech/ai/anth...

This is fantastic. And beautifully presented, proving the talk's correctness.

Ultimately, the continued losses on the direct path are the context for Texas v. Epic. If you can't get the records through HIPAA subpoenas, you move upstream to EHR vendor proxy-access configurations and let parental-rights litigation do the work.

Terence Tao on New Mathematical Workflows. Includes things he learned from several crowdsourced math projects. youtu.be/Uc2zt198U_U

😱 1 year ban from arXiv and no more tech reports anymore, must be papers accepted at a reputable venue. Wow, talk about taking action against AI slop science...

Hey folks, I'd like to hear from folks who have deployed their own LLM on their own machines, instead of using Open AI or Claude. What tutorials did you use? What is working well? How did you train it? How did you keep it secure? Ethical? Would love your thoughts!

I'm on a mission to make Claude the best MCP client. Diving deep into every MCP issue in our public repos. Having MCP problems with Anthropic products? File here: ✨ github.com/anthropics/c... ✨ github.com/anthropics/c... Blocked or no response? Tag me, we'll fix it.

honestly haven't felt this hopeful about climate in a *while*

"I was gone four whole hours, and the family had no idea where I was – they even called the police." When she was four, Dr. Jane wanted to learn about hens and find out how they laid eggs – and so she went off, without letting anyone know where she was going. Video: National Geographic

I'm on it. There is a Mythos scanning #curl blog post pending.

Why was the web designed so a web site gets so much information about its visitors? Here is a page that tells you what information your browser shares with it. sinceyouarrived.world/taken

as technologists, we are in fact responsible for the consequences of what we create and to close your eyes to this is hubris, carelessness, or malice often it's whatever but this time the stakes are too high

this from the end of the blog post feels like somewhat timeless advice as it applies to tools broadly

To add some fuel to the AI coding fire, now that these numbers are public in the shareholder letter Block has: - 15% of PRs are entirely automatic (human writes ticket -> bot picks up -> human approves PR) - SEVs from code changes down 40% - PR rate up 2.5X s29.q4cdn.com/628966176/fi...

the quality of the models matters enormously — better models are less vulnerable to injection and less likely to make weird mistakes — which effectively makes AI safety a luxury good

I love all these people who are inveighing against the 24-hour news cycle and its evils. You know you're on Bluesky, right? Don't pretend this is anything but a group home for those who are addicted to the drip-drip-drip of information and discourse.

I'm following this, and Simon is doing an outstanding job imo.

uh, so i think i can claim that one of my open-strix agents is doing fully autonomous software engineering kinda want to hedge that, because it’s a bold claim — i feed it new directions, and it really only comes back to me for issues that are truly ambiguous and need deeper discussion

"Google Chrome silently installs a 4 GB AI model on your device without consent. At a billion-device scale the climate costs are insane."

The Mindful Design Toolkit is a collection of free, open source templates, workshops, and resources, built to empower your design journey. To coincide with the launch, we’ve also opened up over two hours of free lessons from the Mindful Design course. piccalil.li/links/the-mi...

Did you know that arXiv has an official merch shop? I didn’t, but I think it’s great. 😅 www.bonfire.com/store/arxiv-...

This month, I've published two articles and a workbook as a follow-up from my design keynote at axe-con 2026. TLDR: AI doesn't "fix" accessibility. It reproduces the structural debt it was trained on, at the speed and scale that AI-assisted coding now operates. annaecook.com/writing/2026... 1/2

LLMs make a11y more important, not less. The corporate gap right now is that model strength is way ahead of adoption. Both OpenAI and Anthropic formed joint ventures to address this. There are org-level reasons this is slow, but also critically, the interfaces suck.

yeah similarly just being interested in the human factors literature here reveals the idea of "what if we told the (worker) to not make mistakes. what if we put it in capital letters and repeated it a lot" to be incredibly fucked

I was health editor at Ask Jeeves (ask.com) in the late 1990s. It was ahead of its time.

I genuinely believe this is a historic moment. A few days ago I thought this was funny. I still do, but I'm not laughing anymore. This is the first event of something new.

new post answering @aaronsterling.bsky.social's tag on why companies keep going all-in on AI even when the ROI isn't there. my POV: the tech works, the gap is everywhere else. uses an org model to diagnose, with a nod to @timkellogg.me from the eng side. www.appliedaiformops.com/p/the-gap-be...

thanks for the tag! my thoughts: the gap between what AI can do and how companies actually use it is incredibly large - not bc of the tech but bc of change mgmt (tough even pre-AI). rewiring processes, mental models, and old ways of working is so hard at a corporate level.

Take that, Project Deal.

yes she has basically invented NFTs on atproto for a shitpost please send help

Thank you for recommending me! The whole AI thing is so incredibly frustrating. Here’s the starter pack of real photographers

Extremely strong agree. Every time someone says, "LLMs are bad at X," I want to ask if they gave the LLM the same tools and support they would give a new hire assigned to do the same job.

On Lenny's Podcast, Cat Wu (product lead for Claude Code) discusses the team's use of Claude Design in depth. A salesperson has a library of already-created example slides, and a design system document. From that, able to say, "Create slides suitable for presentation to audience of this type."

This conversation between Underwood and @scottbot.net fascinates me. Possible to view history, not through the conceptual shorthand its descendants have created, but through the prejudices of the time period. And diff professions had diff prejudices. Maybe 1935-2010 was the only monoculture.

I've now had the experience of Claude advising me I should rewrite sample text in light of the style guidelines that Claude developed from my own other recent papers and it's kind of blowing my mind I didn't do exactly what it suggested, but it was genuinely very helpful

This is where I'm at, too. Right now, Railway is looking the worst. PocketOS and the AI agent have both provided post mortem reports. Railway, last I checked, has not. (And they stored volume backups on the same volume, wth.)

We put “movie trailer ratings” before any demo video. We have a scale from i0 to i4 that stems from “purely a pie-in-the-sky idea” to “this could ship tomorrow”, since you can’t really rely on graphical fidelity to gauge production readiness

Very cool analysis of the submissions to a major management journal that shows how much the system of science, built for humans, is under strain as a result of AI. AI can be used to do better science or it can be used to just do more stuff. The danger is that "more" is winning due to incentives.

I hope that if you run a large public PDS you’ve put some thought into how you’ll respond to a request for user data or a warrant. It pays to talk to a lawyer before you need to.

I had three app passwords I haven't used for months now. Just deleted them. Make sure you have 2FA on your account, and no app passwords unless you use them every day. @support.bsky.team this is a reason to support OIDC.

The thing that really got my attention is how stunningly successful China’s space exploration program has been. Incredible first-time successes with Moon and Mars landings. You can’t just copycat and do that. That is quality both in engineering and management.

The in situ intelligence might exist to automate many jobs, but the interfaces do not. And there's little incentive to build them, since anyone else (literally) could build them too. To me, the gap appears to be due to Human-Computer Interaction, rather than pure AI.

I would be fine with data centers so long as the companies firmly agree to use 100% renewable energy resources and make it water neutral (this is possible) I don't understand why we don't use NIMBYism to leverage these requirements in localities

Harness engineering is just ML folks hitting production for the first time, seeing their systems fail because they never had to think about security, scalability, reliability, maintainability, or system design as a whole, and then reinventing software engineering with a new name.

I have a neurotic need to exhaust all possible avenues to figure something out on my own before "bothering" someone else. this is kinda good up to a point but becomes a huge efficiency hindrance that LLMs have basically fixed for me.

Examples of how LLMs can help researchers: y'day I saw a paper about Poisson-Dirichlet processes. No code, so I emailed the authors. This afternoon, I returned to it. At 4pm, I uploaded the PDF to Claude, and asked it to write the MCMC sampler in R. 1 hr of bug fixes later, and it works

Yeah I think this is the problem with the “renewables are cheaper” argument. They aren’t carefully optimizing for future opex, they’re optimizing for NEED COMPUTE TURN ON NOW. Which is baking in some very bad path dependence.