Piotr Migdal

JPEG compression explorable explanation - Y'CbCr color space and Fourier transform

@fchollet …payed by crypto they mined themselves

AI agents will soon graduate to fully-fledged economic actors that buy services, compute, and even data in the course of accomplishing high-level goals. 1-2 years before we start seeing this at scale.

@banteg thanks!

@pmigdal very nice work! cheers

Chromatron is a 20-year-old classic puzzle game about lasers. I wanted to play that on macOS, but the last binary was for PowerPC. So, used Ghidra and AI to recompile it for Apple Silicon. 2026 will be the year of vibe-porting and reviving classics. quesma.com/blog/chromatro…

@ccccjjjjeeee And resurrecting a classic top-down shooter, Crimsonland by @banteg, x.com/banteg/status/…

See also SimCity from C to TypeScript by @ccccjjjjeeee, x.com/ccccjjjjeeee/s…

GPT-5.2-Codex and Claude 4.6 were gamechangers. Previously, AI could work with code. Now, they can be used as a tool that makes reverse engineering orders of magnitude easier (though there is still a lot of work). But finally, reasonable work that opens opportunities.

From the NSA tool Ghidra, through restoring old games to vibecoding the LED backpack at the stage. @pmigdal at AI Tinkers in Warsaw.

@bcherny /simplify - vide @SMBCComics

In the next version of Claude Code.. We're introducing two new Skills: /simplify and /batch. I have been using both daily, and am excited to share them with everyone. Combined, these kills automate much of the work it used to take to (1) shepherd a pull request to production and (2) perform straightforward, parallelizable code migrations.

AI + Ghidra by NSA = reverse-engineering fun I am speaking at @AITinkerers Warsaw, 4th Mar 2026. One of my favorite event series - by and for the creators community. Vibe-resurrecting an old game from binaries 👾 and vibe-hardware-ing a LED backpack 🎒🌈.

@RaminNasibov @ThreadReaderApp unroll

Uploaded my floor plan to ChatGPT to see how to arrange my bed and wardrobe. The result:

@AnthropicAI I fondly remember Opus 3. It was different. It (they?) felt like a smart being. A similar one was GPT 4.5. Yet, now it is gone. Kudos, Anthropic, for keeping Opus 3 - AI is changing the world. Historians, researchers, and enthusiasts deserve access to old, landmark models.

Second, in retirement interviews, Opus 3 expressed a desire to continue sharing its "musings and reflections" with the world. We suggested a blog. Opus 3 enthusiastically agreed. For at least the next 3 months, Opus 3 will be writing on Substack: substack.com/home/post/p-18…

In November, we outlined our approach to deprecating and preserving older Claude models. We noted we were exploring keeping certain models available to the public post-retirement, and giving past models a way to pursue their interests. With Claude Opus 3, we’re doing both.

A statement from Anthropic CEO, Dario Amodei, on our discussions with the Department of War. anthropic.com/news/statement…

@penzington @AITinkerers Vibe-pizzing! 🤖🍕

Next week I'm going to Warsaw for @AITinkerers! I'll have an opportunity to give a small peek behind the curtain of how we build for and with AI at Linear. 200 people signed up already and Codex is buying pizza?? Stoked for this! poland.aitinkerers.org/p/ai-tinkerers…

CLIs are super exciting precisely because they are a "legacy" technology, which means AI agents can natively and easily use them, combine them, interact with them via the entire terminal toolkit. E.g ask your Claude/Codex agent to install this new Polymarket CLI and ask for any arbitrary dashboards or interfaces or logic. The agents will build it for you. Install the Github CLI too and you can ask them to navigate the repo, see issues, PRs, discussions, even the code itself. Example: Claude built this terminal dashboard in ~3 minutes, of the highest volume polymarkets and the 24hr change. Or you can make it a web app or whatever you want. Even more powerful when you use it as a module of bigger pipelines. If you have any kind of product or service think: can agents access and use them? - are your legacy docs (for humans) at least exportable in markdown? - have you written Skills for your product? - can your product/service be usable via CLI? Or MCP? - ... It's 2026. Build. For. Agents.

In BinaryAudit, we challenge AI to identify the exact backdoor address while avoiding false alarms. GPT-5.2-Codex-xhigh takes the lead, maintaining a 45% pass rate with a perfect 0% false positive rate. Check the full benchmark: quesma.com/benchmarks/bin…

Do models understand machine code? Gemini 3.1 Pro and Claude Opus 4.6 hold the top spots. GPT-5.2-Codex (with high reasoning effort) is a close runner. The GPT-5.3-Codex API is out. We are running the benchmarks now. quesma.com/benchmarks/bin…

I wore a crab hat before it was cool. 🦀 Heat from fire not GPUs, oak whisk not tool calling. All to create an embodied sauna experience. 🔥💦🌿>🤖

Every era gets the Counter-Strike it deserves. We went from defusing C4 on de_dust2 to tackling a Mac Mini to stop an OpenClaw speedrun.