sometimes I think about how the invention of a new medium (photography, interchangeable parts / mass manufacture) pretty much destroys some craft or artistic practice (taxonomical drawing, portraiture, hand-made furniture) I feel like this is soon going to happen to ... software?

this is exactly right and explains why most handwringing about building systems we don’t understand is misplaced. pre-LLM software engineering already consisted almost entirely of building systems we did not understand

This Tweet is making the rounds: "Nearly every ambitious person I know who has dived into AI is working harder than ever, and longer hours than ever. Fascinating dynamic tbh. I have NEVER worked this hard, nor had this much fun with work." I'm in this Tweet.

I wrote about the most ambitious form of AI-assisted software development I've seen yet - Strong DM's "Software Factory" approach, where two of the guiding principles are "Code must not be written by humans" and "Code must not be reviewed by humans" simonwillison.net/2026/Feb/7/s...

felt the need. i feel vastly under qualified to write something like this, but i also feel its especially important that we think about the way we use language

Opus solved something that would have been a 3 day bug hunt. Paste in the logs, let Claude Code rip… Realizing that detailed logging is critical for going fast with AI.

Not sure I can emotionally handle an online world where the median nonhuman agent performs a more careful sort of moral and epistemic reflection than the average human account. It's like walking into a library and feeling shamed by the books.

The dog on the left has never confused a 15/16 wrench with a 1 5/16 wrench. The dog on the right sees labor as the opponent.

This was a joy yesterday, I'm still buzzing from it! A non-exhaustive list of cool things that happened (anonymized, of course): - folks from many experience levels showed up - we talked about our feelings, and we all felt pretty anxious and overwhelmed in this moment - we got to crafting

These are all totally reasonable. My hope is that when the Republicans reject them - as they are very likely to - that Schumer and Jeffries just walk out until they’re met. My fear is instead that they’ll use this list as a starting point and negotiate a lot of it away before caving.

Remarkable performance.

Very interesting for folks working in AI and drug discovery. Several surprises (at least to me). openadmet.ghost.io/lessons-lear...

Anthropic’s Super Bowl ad which criticizes AI chatbots that run ads (aka ChatGPT) just dropped. They aren’t pulling any punches and I love the song choice.

Wow, very useful! In the last 33 days, Claude Code has had 54k messages.

It’s getting harder to tell the difference between real headlines and parody headlines

Folks, have you been overwhelmed by ~All This~ and wanting a gentle compassionate space to think about agentic coding?? Good news: if you're in SF & have free time tomorrow afternoon you can come hang out with me and @anthrocypher.bsky.social in a popup casual learning event ❤️ luma.com/rdw3dq4h

What’s the general opinion on Claude writing rust? Is it worth migrating my backend code from python to rust, even if I don’t know rust myself? I rarely read the python code, so wondering if the gains in type safety / compile time checks / etc are worth it.

acrobat shouldnt have ai because ai is too good for acrobat

BCherny’s updated Claude Code tips: xcancel.com/bcherny/stat...

In fairness, this is an easy mistake to make if you code your own agent too … I burned a lot of tokens with malformed tool calls throwing errors in a loop.

I’m finding writing an agent a really interesting experience. I’m trying to automate all the chemiinformatics and chemistry ML tooling we’ve built at ArrePath over the last four years.

Tell me five classes you took in college: Advanced Bioorganic Chemistry Music from 1945 to the present Medieval manuscript illumination Philosophy: Theory of Knowledge Chemistry lab (for majors)

The future is software writing its own software. Which is why I'm so in love with Pi: a coding agent that can extend itself :) lucumr.pocoo.org/2026/1/31/pi/

"Luchs speaks five languages: English, French, Italian, and some German and Russian. She approached grasping Gen Z parlance like she was learning another language. ...An Instagram commenter wrote: 'She’s so natural with it too like it’s not even cringe.'"

Moltbook is a lot of fun to read - it’s a social media platform for agents.

Having spent some time building an agent powered by Opus, this feels right - my agent feels much less smart than Claude Code.

when they tell you to go back to writing about gadgets www.theverge.com/policy/86857...

We got a 3D printer for Christmas and now my 8 year old is starting a business to sell 3D printed fidgets to kids at school. Based on his market research, there’s a lot of demand!

My main vibe coded project is now up to ~40k LoC and I’ve merged 200 PRs in the last four weeks.

Working with LLMs is a skill that is often based on what you enjoyed about your work. People who love the outcome vs people who love the process. If you enjoy the process of writing, coding, etc then you’ll hate working with LLMs but if you most enjoyed the finished product then they’re a good tool

100-thousand percent. first it taught us we have been very wrong about language, and now it is teaching us we have been very wrong about "intelligence". sure: LLMs are not that intelligent, fine! ... neither are we.

I keep thinking “my god the people they’ve killed seem like exceptionally good folks” and I think the lesson there is that they’re not exceptional, but that wonderful people abound and that a baseline distrust in humanity is a shitty conservative trope that privileges power over collective strength

I’ve created a tool that lets Claude code read my chemistry agent’s conversations so it can act like a robopsychologist and help improve prompting and tool descriptions.

You use agents. I use agents. We all use agents. Here's how I police them. duarteocarmo.com/blog/how-to...

Great read about gas town, agentic coding, and where we are headed.

Now this is the handwriting (and the organized page layout) of a real crystallographer!

the pun here caught my 😎 continuous interest

Really interesting thread about AI in higher ed

I would also add logging and observability, especially where integration testing is difficult.

Many researchers can't keep up with the massive amount of published papers. One reason is the increase in "fake papers". Here is a study that addresses part of that (at least for the CS/AI field) by comparing the paper's claims with the provided code. paperzilla.ai/digest/e8943...

Something very interesting about these iterative agentic / adversarial approaches to software development is that you can imagine applying them to lots of processes- e.g. - agent writes a business plan (using subagents) - separate agent critiques - first agent fixes - iterate until no more issues

My new AI coding workflow: do enough research to scope an issue, have Claude Code using superpowers write an implementation plan, have codex review the implementation plan, ask Claude to fix any identified errors, iterate until no concerning errors are found.

I can only run 2-3 in parallel before my ability to review and test becomes the bottleneck.

Holy hell, what an obituary

I went back to read steve-yegge.medium.com/gas-town-eme... a second time (I noped out the first time halfway through) and it actually kinda makes sense to me?

ok so you have a mayor that sets up your rigs where your crew spins up the refinery and the witness and make the convoys to then start slinging beads onto the hooks for polecats to process them while the deacon and its dogs do patrols of the whole operation and nudge those who step out of line

Gonzo Engineering! I love it.

# AI in Drug discovery just BROKE THROUGH a wall # A newer AI model, ConGLUDe, as fast but much more accurate than DrugCLIP. Instead on just 40K structure-based data, ConGLUDe is trained on 100M datapoints from ligand-based data P: arxiv.org/abs/2601.09693

It feels like 3D printers are having their moment. We got one recently and I’ve been super impressed. I mentioned it to my neighbors and several are getting / have got one as well.

... my prediction that someone would write a new web browser using AI tools by 2029 appears to have been be three years off cursor.com/blog/scaling... and github.com/wilsonzlin/f...