<blockquote cite="https://twitter.com/kareldoostrlnck/status/2019477361557926281"><p>When I want to quickly implement a one-off experiment in a part of the codebase I am unfamiliar with, I get codex to do extensive due diligence. Codex explores relevant slack channels, reads related discussions, fetches experimental branches from those discussions, and cherry picks useful changes for my experiment. All of this gets summarized in an extensive set of notes, with links back to where each piece of information was found. Using these notes, codex wires the experiment and makes a bunch of hyperparameter decisions I couldn’t possibly make without much more effort.</p></blockquote><p class="cite">&mdash; <a href="https://twitter.com/kareldoostrlnck/status/2019477361557926281">Karel D&#x27;Oosterlinck</a>, I spent $10,000 to automate my research at OpenAI with Codex</p> <p>Tags: <a href="https://simonwillison.net/tags/codex-cli">codex-cli</a>, <a href="https://simonwillison.net/tags/coding-agen

<p><strong><a href="https://mitchellh.com/writing/my-ai-adoption-journey">Mitchell Hashimoto: My AI Adoption Journey</a></strong></p>Some really good and unconventional tips in here for getting to a place with coding agents where they demonstrably improve your workflow and productivity. I particularly liked:</p><ul><li><p><a href="https://mitchellh.com/writing/my-ai-adoption-journey#step-2-reproduce-your-own-work">Reproduce your own work</a> - when learning to use coding agents Mitchell went through a period of doing the work manually, then recreating the same solution using agents as an exercise:</p><blockquote><p>I literally did the work twice. I'd do the work manually, and then I'd fight an agent to produce identical results in terms of quality and function (without it being able to see my manual solution, of course).</p></blockquote></li><li><p><a href="https://mitchellh.com/writing/my-ai-adoption-journey#step-3-end-of-day-agents">End-of-day agents</a> - letting agents step in whe

<p>Two major new model releases today, within about 15 minutes of each other.</p><p>Anthropic <a href="https://www.anthropic.com/news/claude-opus-4-6">released Opus 4.6</a>. Here's <a href="https://gist.github.com/simonw/a6806ce41b4c721e240a4548ecdbe216">its pelican</a>:</p><p><img alt="Slightly wonky bicycle frame but an excellent pelican, very clear beak and pouch, nice feathers." src="https://static.simonwillison.net/static/2026/opus-4.6-pelican.png" /></p><p>OpenAI <a href="https://openai.com/index/introducing-gpt-5-3-codex/">release GPT-5.3-Codex</a>, albeit only via their Codex app, not yet in their API. Here's <a href="https://gist.github.com/simonw/bfc4a83f588ac762c773679c0d1e034b">its pelican</a>:</p><p><img alt="Not nearly as good - the bicycle is a bit mangled, the pelican not nearly as well rendered - it's more of a line drawing." src="https://static.simonwillison.net/static/2026/codex-5.3-pelican.png" /></p><p>I've had a bit of preview access to both of these models and t

<p><strong><a href="https://www.cia.gov/stories/story/spotlighting-the-world-factbook-as-we-bid-a-fond-farewell/">Spotlighting The World Factbook as We Bid a Fond Farewell</a></strong></p>Somewhat devastating news today from CIA:</p><blockquote><p>One of CIA’s oldest and most recognizable intelligence publications, The World Factbook, has sunset.</p></blockquote><p>There's not even a hint as to <em>why</em> they decided to stop maintaining this publication, which has been their most useful public-facing initiative since 1971 and a cornerstone of the public internet since 1997.</p><p>In a bizarre act of cultural vandalism they've not just removed the entire site (including the archives of previous versions) but they've also set every single page to be a 302 redirect to their closure announcement.</p><p>The Factbook has been released into the public domain since the start. There's no reason not to continue to serve archived versions - a banner at the top of the page saying it's no longe

<p><strong><a href="https://mistral.ai/news/voxtral-transcribe-2">Voxtral transcribes at the speed of sound</a></strong></p>Mistral just released Voxtral Transcribe 2 - a family of two new models, one open weights, for transcribing audio to text. This is the latest in their Whisper-like model family, and a sequel to the original Voxtral which they released <a href="https://simonwillison.net/2025/Jul/16/voxtral/">in July 2025</a>.</p><p>Voxtral Realtime - official name <code>Voxtral-Mini-4B-Realtime-2602</code> - is the open weights (Apache-2.0) model, available as a <a href="https://huggingface.co/mistralai/Voxtral-Mini-4B-Realtime-2602">8.87GB download from Hugging Face</a>.</p><p>You can try it out in this <a href="https://huggingface.co/spaces/mistralai/Voxtral-Mini-Realtime">live demo</a> - don't be put off by the "No microphone found" message, clicking "Record" should have your browser request permission and then start the demo working. I was very impressed by the demo - I talked

<p>I've been exploring Go for building small, fast and self-contained binary applications recently. I'm enjoying how there's generally one obvious way to do things and the resulting code is boring and readable - and something that LLMs are very competent at writing. The one catch is distribution, but it turns out publishing Go binaries to PyPI means any Go binary can be just a <code>uvx package-name</code> call away.</p><h4 id="sqlite-scanner">sqlite-scanner</h4><p><a href="https://github.com/simonw/sqlite-scanner">sqlite-scanner</a> is my new Go CLI tool for scanning a filesystem for SQLite database files.</p><p>It works by checking if the first 16 bytes of the file exactly match the SQLite magic number sequence <code>SQLite format 3\x00</code>. It can search one or more folders recursively, spinning up concurrent goroutines to accelerate the scan. It streams out results as it finds them in plain text, JSON or newline-delimited JSON. It can optionally display the file sizes as well.<

<p><strong><a href="https://deno.com/blog/introducing-deno-sandbox">Introducing Deno Sandbox</a></strong></p>Here's a new hosted sandbox product from the Deno team. It's actually unrelated to Deno itself - this is part of their Deno Deploy SaaS platform. As such, you don't even need to use JavaScript to access it - you can create and execute code in a hosted sandbox using their <a href="https://pypi.org/project/deno-sandbox/">deno-sandbox</a> Python library like this:</p><div class="highlight highlight-source-shell"><pre><span class="pl-k">export</span> DENO_DEPLOY_TOKEN=<span class="pl-s"><span class="pl-pds">"</span>... API token ...<span class="pl-pds">"</span></span>uv run --with deno-sandbox python</pre></div><p>Then:</p><pre><span class="pl-k">from</span> <span class="pl-s1">deno_sandbox</span> <span class="pl-k">import</span> <span class="pl-v">DenoDeploy</span><span class="pl-s1">sdk</span> <span class="pl-c1">=</span> <span class="pl-en">DenoDeploy</span>()<span class="pl-k">

<p>I just sent the January edition of my <a href="https://github.com/sponsors/simonw/">sponsors-only monthly newsletter</a>. If you are a sponsor (or if you start a sponsorship now) you can <a href="https://github.com/simonw-private/monthly/blob/main/2026-01-january.md">access it here</a>. In the newsletter for January:</p><ul><li>LLM predictions for 2026</li><li>Coding agents get even more attention</li><li>Clawdbot/Moltbot/OpenClaw went very viral</li><li>Kakapo breeding season is off to a really strong start</li><li>New options for sandboxes</li><li>Web browsers are the "hello world" of coding agent swarms</li><li>Sam Altman addressed the Jevons paradox for software engineering</li><li>Model releases and miscellaneous extras</li></ul><p>Here's <a href="https://gist.github.com/simonw/13e595a236218afce002e9aeafd75cd0">a copy of the December newsletter</a> as a preview of what you'll get. Pay $10/month to stay a month ahead of the free copy!</p> <p>Tags: <a href="https://simonwillison

<blockquote cite="https://www.youtube.com/watch?v=mb3uK-_QkOo&amp;t=832s"><p>This is the difference between Data and a large language model, at least the ones operating right now. Data created art because he wanted to grow. He wanted to become something. He wanted to understand. Art is the means by which we become what we want to be. [...]</p><p>The book, the painting, the film script is not the only art. It's important, but in a way it's a receipt. It's a diploma. The book you write, the painting you create, the music you compose is important and artistic, but it's also a mark of proof that you have done the work to learn, because in the end of it all, you are the art. The most important change made by an artistic endeavor is the change it makes in you. The most important emotions are the ones you feel when writing that story and holding the completed work. I don't care if the AI can create something that is better than what we can create, because it cannot be changed by that creatio

<p><strong><a href="https://openai.com/index/introducing-the-codex-app/">Introducing the Codex app</a></strong></p>OpenAI just released a new macOS app for their Codex coding agent. I've had a few days of preview access - it's a solid app that provides a nice UI over the capabilities of the Codex CLI agent and adds some interesting new features, most notably first-class support for <a href="https://developers.openai.com/codex/skills">Skills</a>, and <a href="https://developers.openai.com/codex/app/automations">Automations</a> for running scheduled tasks.</p><p><img alt="Screenshot of a macOS desktop application with a dark sidebar and light main content area. Left sidebar shows navigation items &quot;New thread&quot;, &quot;Automations&quot;, &quot;Skills&quot;, and a &quot;Threads&quot; section containing two project folders: &quot;local-codex-scratch&quot; with tasks &quot;Reply to greeting task 2h&quot; and &quot;List Codex.app contents 3h&quot;, and &quot;shot-scraper&quot; with t

<p><strong><a href="https://www.nytimes.com/2026/02/02/technology/moltbook-ai-social-media.html?unlocked_article_code=1.JFA.kBCd.hUw-s4vvfswK&amp;smid=url-share">A Social Network for A.I. Bots Only. No Humans Allowed.</a></strong></p>I talked to Cade Metz for this New York Times piece on OpenClaw and Moltbook. Cade reached out after seeing my <a href="https://simonwillison.net/2026/Jan/30/moltbook/">blog post about that</a> from the other day.</p><p>In a first for me, they decided to send a photographer, Jason Henry, to my home to take some photos for the piece! That's my grubby laptop screen at the top of the story (showing <a href="https://www.moltbook.com/post/6e8c3a2c-5f9f-44bc-85ef-770a8d605598">this post</a> on Moltbook). There's a photo of me later in the story too, though sadly not one of the ones that Jason took that included our chickens.</p><p>Here's my snippet from the article:</p><blockquote><p>He was entertained by the way the bots coaxed each other into talking like mac

<p><strong><a href="https://til.simonwillison.net/llms/openclaw-docker">TIL: Running OpenClaw in Docker</a></strong></p>I've been running <a href="https://openclaw.ai/">OpenClaw</a> using Docker on my Mac. Here are the first in my ongoing notes on how I set that up and the commands I'm using to administer it.</p><ul><li><a href="https://til.simonwillison.net/llms/openclaw-docker#use-their-docker-compose-configuration">Use their Docker Compose configuration</a></li><li><a href="https://til.simonwillison.net/llms/openclaw-docker#answering-all-of-those-questions">Answering all of those questions</a></li><li><a href="https://til.simonwillison.net/llms/openclaw-docker#running-administrative-commands">Running administrative commands</a></li><li><a href="https://til.simonwillison.net/llms/openclaw-docker#setting-up-a-telegram-bot">Setting up a Telegram bot</a></li><li><a href="https://til.simonwillison.net/llms/openclaw-docker#accessing-the-web-ui">Accessing the web UI</a></li><li><a href="h

<blockquote cite="https://twitter.com/karpathy/status/2017703360393318587"><p>Originally in 2019, GPT-2 was trained by OpenAI on 32 TPU v3 chips for 168 hours (7 days), with $8/hour/TPUv3 back then, for a total cost of approx. $43K. It achieves 0.256525 CORE score, which is an ensemble metric introduced in the DCLM paper over 22 evaluations like ARC/MMLU/etc.</p><p>As of the last few improvements merged into nanochat (many of them originating in modded-nanogpt repo), I can now reach a higher CORE score in 3.04 hours (~$73) on a single 8XH100 node. This is a 600X cost reduction over 7 years, i.e. the cost to train GPT-2 is falling approximately 2.5X every year.</p></blockquote><p class="cite">&mdash; <a href="https://twitter.com/karpathy/status/2017703360393318587">Andrej Karpathy</a></p> <p>Tags: <a href="https://simonwillison.net/tags/andrej-karpathy">andrej-karpathy</a>, <a href="https://simonwillison.net/tags/gpt-2">gpt-2</a>, <a href="https://simonwillison.net/tags/generative-ai">

<p><strong><a href="https://interconnected.org/home/2026/01/30/efficacy">Singing the gospel of collective efficacy</a></strong></p>Lovely piece from Matt Webb about how you can "just do things" to help make your community better for everyone:</p><blockquote><p>Similarly we all love when the swifts visit (beautiful birds), so somebody started a group to get swift nest boxes made and installed collectively, then applied for subsidy funding, then got everyone to chip in such that people who couldn’t afford it could have their boxes paid for, and now suddenly we’re all writing to MPs and following the legislation to include swift nesting sites in new build houses. Etc.</p><p>It’s called <em>collective efficacy</em>, the belief that you can make a difference by acting together.</p></blockquote><p>My current favorite "you can just do things" is a bit of a stretch, but apparently you can just build a successful software company for 20 years and then use the proceeds to <a href="https://bmore

<blockquote cite="https://steve-yegge.medium.com/software-survival-3-0-97a2a6255f7b"><p>Getting agents using Beads requires much less prompting, because Beads now has 4 months of “Desire Paths” design, which I’ve talked about before. Beads has evolved a very complex command-line interface, with 100+ subcommands, each with many sub-subcommands, aliases, alternate syntaxes, and other affordances.</p><p>The complicated Beads CLI isn’t for humans; it’s for agents. What I did was make their hallucinations real, over and over, by implementing whatever I saw the agents trying to do with Beads, until nearly every guess by an agent is now correct.</p></blockquote><p class="cite">&mdash; <a href="https://steve-yegge.medium.com/software-survival-3-0-97a2a6255f7b">Steve Yegge</a>, Software Survival 3.0</p> <p>Tags: <a href="https://simonwillison.net/tags/steve-yegge">steve-yegge</a>, <a href="https://simonwillison.net/tags/coding-agents">coding-agents</a>, <a href="https://simonwillison.net/tags/

<p>The hottest project in AI right now is Clawdbot, <a href="https://x.com/openclaw/status/2016058924403753024">renamed to Moltbot</a>, <a href="https://openclaw.ai/blog/introducing-openclaw">renamed to OpenClaw</a>. It's an open source implementation of the digital personal assistant pattern, built by Peter Steinberger to integrate with the messaging system of your choice. It's two months old, has over 114,000 stars <a href="https://github.com/openclaw/openclaw">on GitHub</a> and is seeing incredible adoption, especially given the friction involved in setting it up.</p><p>(Given the <a href="https://x.com/rahulsood/status/2015397582105969106">inherent risk of prompt injection</a> against this class of software it's my current pick for <a href="https://simonwillison.net/2026/Jan/8/llm-predictions-for-2026/#1-year-a-challenger-disaster-for-coding-agent-security">most likely to result in a Challenger disaster</a>, but I'm going to put that aside for the moment.)</p><p>OpenClaw is built

<p><strong><a href="https://www.tiktok.com/@chris_ashworth/video/7600801037292768525">We gotta talk about AI as a programming tool for the arts</a></strong></p>Chris Ashworth is the creator and CEO of <a href="https://en.wikipedia.org/wiki/QLab">QLab</a>, a macOS software package for “cue-based, multimedia playback” which is designed to automate lighting and audio for live theater productions.</p><p>I recently started following him on TikTok where he posts about his business and theater automation in general - Chris founded <a href="https://voxel.org/faq/">the Voxel</a> theater in Baltimore which QLab use as a combined performance venue, teaching hub and research lab (here's <a href="https://bmoreart.com/2024/09/the-voxel-is-a-cutting-edge-theater-experiment.html">a profile of the theater</a>), and the resulting videos offer a fascinating glimpse into a world I know virtually nothing about.</p><p><a href="https://www.tiktok.com/@chris_ashworth/video/7600801037292768525">This latest Ti

<p><strong><a href="https://docs.datasette.io/en/latest/changelog.html#a24-2026-01-29">Datasette 1.0a24</a></strong></p>New Datasette alpha this morning. Key new features:</p><ul><li>Datasette's <code>Request</code> object can now handle <code>multipart/form-data</code> file uploads via the new <a href="https://docs.datasette.io/en/latest/internals.html#internals-formdata">await request.form(files=True)</a> method. I plan to use this for a <code>datasette-files</code> plugin to support attaching files to rows of data.</li><li>The <a href="https://docs.datasette.io/en/latest/contributing.html#setting-up-a-development-environment">recommended development environment</a> for hacking on Datasette itself now uses <a href="https://github.com/astral-sh/uv">uv</a>. Crucially, you can clone Datasette and run <code>uv run pytest</code> to run the tests without needing to manually create a virtual environment or install dependencies first, thanks to the <a href="https://til.simonwillison.net/uv/

<p>My blog uses aggressive caching: it sits behind Cloudflare with a 15 minute cache header, which guarantees it can survive even the largest traffic spike to any given page. I've recently added a couple of dynamic features that work in spite of that full-page caching. Here's how those work.</p><h4 id="edit-links-that-are-visible-only-to-me">Edit links that are visible only to me</h4><p>This is a Django site and I manage it through the Django admin.</p><p>I have <a href="https://github.com/simonw/simonwillisonblog/blob/b8066f870a94d149f5e8cee6e787d3377c0b9507/blog/models.py#L254-L449">four types of content</a> - entries, link posts (aka blogmarks), quotations and notes. Each of those has a different model and hence a different Django admin area.</p><p>I wanted an "edit" link on the public pages that was only visible to me.</p><p>The button looks like this:</p><p><img src="https://static.simonwillison.net/static/2026/edit-link.jpg" alt="Entry footer - it says Posted 27th January 2026 a

<p><strong><a href="https://www.danshapiro.com/blog/2026/01/the-five-levels-from-spicy-autocomplete-to-the-software-factory/">The Five Levels: from Spicy Autocomplete to the Dark Factory</a></strong></p>Dan Shapiro proposes a five level model of AI-assisted programming, inspired by the five (or rather six, it's zero-indexed) <a href="https://www.nhtsa.gov/sites/nhtsa.gov/files/2022-05/Level-of-Automation-052522-tag.pdf">levels of driving automation</a>.</p><ol start="0"><li><strong>Spicy autocomplete</strong>, aka original GitHub Copilot or copying and pasting snippets from ChatGPT.</li><li>The <strong>coding intern</strong>, writing unimportant snippets and boilerplate with full human review.</li><li>The <strong>junior developer</strong>, pair programming with the model but still reviewing every line.</li><li>The <strong>developer</strong>. Most code is generated by AI, and you take on the role of full-time code reviewer.</li><li>The <strong>engineering team</strong>. You're more of

<p><strong><a href="https://emsh.cat/one-human-one-agent-one-browser/">One Human + One Agent = One Browser From Scratch</a></strong></p>embedding-shapes was <a href="https://emsh.cat/cursor-implied-success-without-evidence/">so infuriated</a> by the hype around Cursor's <a href="https://simonwillison.net/2026/Jan/23/fastrender/">FastRender browser project</a> - thousands of parallel agents producing ~1.6 million lines of Rust - that they were inspired to take a go at building a web browser using coding agents themselves.</p><p>The result is <a href="https://github.com/embedding-shapes/one-agent-one-browser">one-agent-one-browser</a> and it's <em>really</em> impressive. Over three days they drove a single Codex CLI agent to build 20,000 lines of Rust that successfully renders HTML+CSS with no Rust crate dependencies at all - though it does (reasonably) use Windows, macOS and Linux system frameworks for image and text rendering.</p><p>I installed the <a href="https://github.com/embeddin

<p><strong><a href="https://www.kimi.com/blog/kimi-k2-5.html">Kimi K2.5: Visual Agentic Intelligence</a></strong></p>Kimi K2 landed <a href="https://simonwillison.net/2025/Jul/11/kimi-k2/">in July</a> as a 1 trillion parameter open weight LLM. It was joined by Kimi K2 Thinking <a href="https://simonwillison.net/2025/Nov/6/kimi-k2-thinking/">in November</a> which added reasoning capabilities. Now they've made it multi-modal: the K2 models were text-only, but the new 2.5 can handle image inputs as well:</p><blockquote><p>Kimi K2.5 builds on Kimi K2 with continued pretraining over approximately 15T mixed visual and text tokens. Built as a native multimodal model, K2.5 delivers state-of-the-art coding and vision capabilities and a self-directed agent swarm paradigm.</p></blockquote><p>The "self-directed agent swarm paradigm" claim there means improved long-sequence tool calling and training on how to break down tasks for multiple agents to work on at once:</p><blockquote><p>For complex ta

<p>Someone <a href="https://news.ycombinator.com/item?id=46765460#46765823">asked</a> on Hacker News if I had any tips for getting coding agents to write decent quality tests. Here's what I said:</p><hr /><p>I work in Python which helps a lot because there are a TON of good examples of pytest tests floating around in the training data, including things like usage of fixture libraries for mocking external HTTP APIs and snapshot testing and other neat patterns.</p><p>Or I can say "use pytest-httpx to mock the endpoints" and Claude knows what I mean.</p><p>Keeping an eye on the tests is important. The most common anti-pattern I see is large amounts of duplicated test setup code - which isn't a huge deal, I'm much more more tolerant of duplicated logic in tests than I am in implementation, but it's still worth pushing back on.</p><p>"Refactor those tests to use pytest.mark.parametrize" and "extract the common setup into a pytest fixture" work really well there.</p><p>Generally though the

<p>One of my favourite features of ChatGPT is its ability to write and execute code in a container. This feature launched as ChatGPT Code Interpreter <a href="https://simonwillison.net/2023/Apr/12/code-interpreter/">nearly three years ago</a>, was half-heartedly rebranded to "Advanced Data Analysis" at some point and is generally really difficult to find detailed documentation about. Case in point: it appears to have had a <em>massive</em> upgrade at some point in the past few months, and I can't find documentation about the new capabilities anywhere!</p><p>Here are the most notable new features:</p><ol><li>ChatGPT can <strong>directly run Bash commands</strong> now. Previously it was limited to Python code only, although it could run shell commands via the Python <code>subprocess</code> module.</li><li><strong>It has Node.js</strong> and can run JavaScript directly in addition to Python. I also got it to run "hello world" in <strong>Ruby, Perl, PHP, Go, Java, Swift, Kotlin, C and C++

<p><strong><a href="https://aifoc.us/the-browser-is-the-sandbox/">the browser is the sandbox</a></strong></p>Paul Kinlan is a web platform developer advocate at Google and recently turned his attention to coding agents. He quickly identified the importance of a robust sandbox for agents to operate in and put together these detailed notes on how the web browser can help:</p><blockquote><p>This got me thinking about the browser. Over the last 30 years, we have built a sandbox specifically designed to run incredibly hostile, untrusted code from anywhere on the web, the instant a user taps a URL. [...]</p><p>Could you build something like Cowork in the browser? Maybe. To find out, I built a demo called <a href="http://co-do.xyz">Co-do</a> that tests this hypothesis. In this post I want to discuss the research I've done to see how far we can get, and determine if the browser's ability to run untrusted code is useful (and good enough) for enabling software to do more for us directly on our

<p><strong><a href="https://www.doc.govt.nz/our-work/kakapo-recovery/what-we-do/kakapo-cam-rakiura-live-stream/">Kākāpō Cam: Rakiura live stream</a></strong></p>Critical update for this year's <a href="https://simonwillison.net/2026/Jan/8/llm-predictions-for-2026/#1-year-k-k-p-parrots-will-have-an-outstanding-breeding-season">Kākāpō breeding season</a>: the New Zealand Department of Conservation have a livestream running of Rakiura's nest!</p><blockquote><p>You’re looking at the underground nest of 23-year-old Rakiura. She has chosen this same site to nest for all seven breeding seasons since 2008, a large cavity under a rātā tree. Because she returns to the site so reliably, we’ve been able to make modifications over the years to keep it safe and dry, including adding a well-placed hatch for monitoring eggs and chicks.</p></blockquote><p>Rakiura is a legendary Kākāpō:</p><blockquote><p>Rakiura hatched on 19 February 2002 on Whenua Hou/Codfish Island. She is the offspring of Flossie a

<p><strong><a href="https://www.youtube.com/watch?v=4u94juYwLLM">Don&#x27;t &quot;Trust the Process&quot;</a></strong></p>Jenny Wen, Design Lead at Anthropic (and previously Director of Design at Figma) gave a provocative keynote at Hatch Conference in Berlin last September.</p><p><img alt="Don't &quot;Trust the process&quot; slide, speaker shown on the left" src="https://static.simonwillison.net/static/2026/dont-trust-process.jpg" /></p><p>Jenny argues that the Design Process - user research leading to personas leading to user journeys leading to wireframes... all before anything gets built - may be outdated for today's world.</p><blockquote><p><strong>Hypothesis</strong>: In a world where anyone can make anything — what matters is your ability to choose and curate what you make.</p></blockquote><p>In place of the Process, designers should lean into prototypes. AI makes these much more accessible and less time-consuming than they used to be.</p><p>Watching this talk made me think abo

<blockquote cite="https://jasmi.news/p/claude-code"><p><strong>If you tell a friend they can now instantly create any app, they’ll probably say “Cool! Now I need to think of an idea.”</strong> Then they will forget about it, and never build a thing. The problem is not that your friend is horribly uncreative. It’s that most people’s problems are not software-shaped, and most won’t notice even when they are. [...]</p><p>Programmers are trained to see everything as a software-shaped problem: if you do a task three times, you should probably automate it with a script. <em>Rename every IMG_*.jpg file from the last week to hawaii2025_*.jpg</em>, they tell their terminal, while the rest of us painfully click and copy-paste. We are blind to the solutions we were never taught to see, asking for faster horses and never dreaming of cars.</p></blockquote><p class="cite">&mdash; <a href="https://jasmi.news/p/claude-code">Jasmine Sun</a></p> <p>Tags: <a href="https://simonwillison.net/tags/vibe-cod

<p>Last week Cursor published <a href="https://cursor.com/blog/scaling-agents">Scaling long-running autonomous coding</a>, an article describing their research efforts into coordinating large numbers of autonomous coding agents. One of the projects mentioned in the article was <a href="https://github.com/wilsonzlin/fastrender">FastRender</a>, a web browser they built from scratch using their agent swarms. I wanted to learn more so I asked Wilson Lin, the engineer behind FastRender, if we could record a conversation about the project. That 47 minute video is <a href="https://www.youtube.com/watch?v=bKrAcTf2pL4">now available on YouTube</a>. I've included some of the highlights below.</p><iframe style="margin-top: 1.5em; margin-bottom: 1.5em;" width="560" height="315" src="https://www.youtube-nocookie.com/embed/bKrAcTf2pL4" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" referrerpoli

<blockquote cite="https://twitter.com/voooooogel/status/2014189072647078053"><p>[...] i was too busy with work to read anything, so i asked chatgpt to summarize some books on state formation, and it suggested circumscription theory. there was already the natural boundary of my computer hemming the towns in, and town mayors played the role of big men to drive conflict. so i just needed a way for them to fight. i slightly tweaked the allocation of claude max accounts to the towns from a demand-based to a fixed allocation system. towns would each get a fixed amount of tokens to start, but i added a soldier role that could attack and defend in raids to steal tokens from other towns. [...]</p></blockquote><p class="cite">&mdash; <a href="https://twitter.com/voooooogel/status/2014189072647078053">Theia Vogel</a>, Gas Town fan fiction</p> <p>Tags: <a href="https://simonwillison.net/tags/parallel-agents">parallel-agents</a>, <a href="https://simonwillison.net/tags/llms">llms</a>, <a href="htt