JSC retweeted
who in SF is building fun hardware with OpenAI APIs and Codex?
think fidget toys, little assistants, weird gadgets
i’d love to capture your use cases and chat about what you’re making
JSC retweeted
Most people fail not because they make the wrong move, but because they never make the first one.
Go try the idea. Fall flat on your face if you have to. The cost of inaction is almost always higher than the cost of a failed attempt.
JSC retweeted
A Monday morning question for you:
Let's say it doesn't work out for some reason. What is something that is still worth trying because of what you will learn along the way?
JSC retweeted
First recording of the day with @MachJustine from @browserbase.
We had the hottest AI Takes at @aiDotEngineer 🔥🔥
JSC retweeted
Take a look at your favourite skill.
Go on, take a look.
Check for lines like:
- "Make the commit message very detailed"
- "Be thorough"
- "Make the implementation easy to read"
What do these lines have in common?
They're no-ops. They do nothing to change the agent's behavior.
Agents always write good commit messages. They always try to be thorough. They always try to create easy-to-read implementations.
Try removing them. Does the output change? No? Then the line was a no-op.
Agent-authored skills are LITTERED with no-ops. Entire paragraphs and sections of text that does nothing to change the agents' behavior.
No-ops make skills harder to evaluate, harder to maintain, and burn tokens uselessly.
Learn to spot them, test them, and remove them. Your skills will thank you.
JSC retweeted
Two months ago I was fired by Google for creating the Google Workspace CLI. It went viral, hit #1 on Hacker News, gained thousands of GitHub stars and many thousands of actual users in just a couple days.
It was an incredible, confusing journey, from directors and leaders asking what they could learn from the tool to getting grilled by legal about why the Google logo and brand colors are on the Google Workspace GitHub code repositories.
I think the cause was that Workspace and certain leaders (and projects) were afraid of being disrupted. But the fear wasn't specific to my CLI, it was a broader fear in what agents meant for Workspace. Either way, the irony of my termination was the announcement at Google Cloud Next two days before I was fired that an official Workspace CLI was coming.
I want this out there because it is easier for me to explain my story and it is an experience I want to fully own. It's also part of my healing.
Nearly 7 years at Google was an incredible opportunity for me and I was fortunate to have wonderful teammates and a manager that fully supported me through these last few months. Thank you.
JSC retweeted
The next hot programming language is… markdown.
A minimal eve agent:
📂 𝚊𝚐𝚎𝚗𝚝/
📄 𝚒𝚗𝚜𝚝𝚛𝚞𝚌𝚝𝚒𝚘𝚗𝚜.𝚖𝚍
📂 𝚜𝚔𝚒𝚕𝚕𝚜/
📄 𝚢𝚘𝚞𝚛-𝚎𝚡𝚙𝚎𝚛𝚝𝚒𝚜𝚎.𝚖𝚍
Deployable in one command: 𝚟𝚎𝚛𝚌𝚎𝚕.
It’s the most accessible programming has ever been. And likely will ever be, at least for the generation of software fully defined and controlled by us humans.
(As a fun fact, one of the initial prototypes for eve was codenamed 𝚕𝚊𝚜𝚝 by @timolins, both in homage to ‘@nextjs for agents’ but also in recognition of how enduring eve’s design feels to us.)
JSC retweeted
This is either brilliant or scary:
Anthropic accidentally leaked the TS source code of Claude Code (which is closed source). Repos sharing the source are taken down with DMCA.
BUT this repo rewrote the code using Python, and so it violates no copyright & cannot be taken down!
JSC retweeted
Most beautiful code I have seen shared in public recently.
Built by Andrej Karpathy - single file of 200 lines of pure Python with no dependencies that trains and inferences a GPT. This is how it should be taught to everyone trying to get into learning LLMs.
This might be the cleanest, most elegant public code drop in AI this year.
Karpathy's new "art project": microgpt (karpathy.github.io/2026/02/1…)
→ Single Python file (~200 lines)
→ No PyTorch, no NumPy, no external libraries at all
→ Full working GPT: data loading → character tokenizer → tiny autograd engine → GPT-2-style transformer → Adam optimizer → training loop → inference/sampling
It's the bare-metal essence of what makes large language models tick - everything else (CUDA kernels, distributed training, mixed precision, flash attention, massive datasets…) is optimization & engineering around this core.
Perfect starting point for anyone trying to truly understand LLMs instead of just calling APIs.
Highly recommend reading + running it. Changes how you see "AI is just matrix multiplies + softmax" from abstract → concrete.
JSC retweeted
exploring shapes of thoughts: extracted my obsidian notes' embeddings and arranges them as a 3d network using 3 different topologies:
- centralized: one core idea connecting all
- decentralized: notes cluster into themed hubs
- distributed: edges labeled by llm describing how ideas connect
JSC retweeted
Someone has to prompt the Claudes, talk to customers, coordinate with other teams, decide what to build next. Engineering is changing and great engineers are more important than ever.
JSC retweeted
Excited for this! I think it is by far the most interesting approach of any BCI effort.
Introducing @merge: Bridging biological and artificial intelligence to maximize human ability, agency and experience.
We’re starting out as a research lab, with our ultimate measure of success being products that people love.
Consider joining us!
JSC retweeted
Bad AGENTS.md files can make your coding agent worse and cost you tokens.
Here's a prompt you can use to clean them up - PLUS a full guide for folks wanting to learn more.
Enjoy: aihero.dev/s/KQLczQ
JSC retweeted
Wrote a bit more about this in my blog simonwillison.net/2026/Jan/1…
JSC retweeted
My talk from the @aiDotEngineer Code Summit is out! 🚨
"How Claude Code Works" and what we can learn about frontier agent architectures.
Coding agents are suddenly really really good, and I'm trying to understand why. In short: better models, simple loop design, and bash tools
JSC retweeted
When I created Claude Code as a side project back in September 2024, I had no idea it would grow to be what it is today. It is humbling to see how Claude Code has become a core dev tool for so many engineers, how enthusiastic the community is, and how people are using it for all sorts of things from coding, to devops, to research, to non-technical use cases. This technology is alien and magical, and it makes it so much easier for people to build and create. Increasingly, code is no longer the bottleneck.
A year ago, Claude struggled to generate bash commands without escaping issues. It worked for seconds or minutes at a time. We saw early signs that it may become broadly useful for coding one day.
Fast forward to today. In the last thirty days, I landed 259 PRs -- 497 commits, 40k lines added, 38k lines removed. Every single line was written by Claude Code + Opus 4.5. Claude consistently runs for minutes, hours, and days at a time (using Stop hooks). Software engineering is changing, and we are entering a new period in coding history. And we're still just getting started..
