a thread with some of my top animations 📍🧵 https://t.co/kyAsrwERAT

I made a Blender animation showing various nice ways to inscribe a Platonic solid inside another. Icosahedron -> Octahedron -> Tetrahedron -> Cube -> Dodecahedron -> Icosahedron

See our latest product being featured on KRON4 San Francisco: Bluelene Fit+Firm Serum to counteract GLP-1 Ozempic Face

‘Sovereign AI’ means fine tuning a model from an American company on data translated to Welsh by models from a Chinese company and another American company?

Introducing Reka Speech: an efficient and accurate transcription & translation model. 🗣️ On modern GPUs (e.g. H100), it runs 8x–35x faster than existing solutions for batch processing.

@jxmnop Their lora training examples are just broken- https://t.co/zvMssbykgI Even if it worked, you wouldnt get the full efficiency because we can’t merge the loras into the quantized layers And even if we could, it only works on h100+

Introducing XRPR: The first U.S. ETF giving you spot exposure to XRP via a traditional ETF.

🚨 New benchmark release 🚨 We're introducing Research-Eval: a diverse, high-quality benchmark for evaluating search-augmented LLMs 👉 Blogpost: https://t.co/fLJonKsJ08 👉 Dataset + code: https://t.co/fJiySwgO4s 🧵 Here's why this matters:

has anyone post-trained 7b-ish models to be better at tokenization awareness? Finding cases where they make quite bad mistakes when they need to manipulate strings in simple ways

@jeremyphoward writing mcp services for accessing your software's documentation is an interesting pattern- e.g. how @Vapi_AI does it

I created one of these logos in Keynote ®

built with

run this yourself by grabbing WordLadder.jsx from https://t.co/TAGljFY53D and running leafstone WordLadder.jsx

i got my computer solving endless word ladder puzzles depth first search, but prioritising words that get you closer to the goal. what algorithm would you use?

We got a glimpse of Playoff Randy last night Presented by @NutrafolMen

https://t.co/5pfNFVVCft

try it out and let me know what you think! npm install -g leafstone

when you're ready to share: leafstone --build ./dist Component.jsx Outputs static files you can deploy anywhere

You can use any external libraries by adding a comment: // @requires recharts@^2.8.0 leafstone sees these and auto-installs. Same for assets: // @requires-asset ./chart-data.json

Year-end $SPX predictions..

I want to reduce the friction for playing with new dynamic components, animations, etc. you don't need to set up a react project, configure vite, set up package.json, configure tailwind etc- just create a new file and point leafstone at it

Zero config needed. Point leafstone at any .jsx file and get: • Instant hot reloading • Tailwind CSS + Lucide Icons built-in (just like Claude and ChatGPT assume) • Auto-installs dependencies Perfect for AI-generated components.

ChatGPT Canvas and Claude Artifacts both landed on the same pattern for standalone, interactive UIs: single-file React components. but how do you actually run a standalone React component? Today I'm releasing leafstone run `leafstone MyComponent.jsx` for an instant dev server

Is there anything like llms.txt for software (rather than websites)? e.g. ‘my-cli-tool --llm-help’ or ‘python -m mypackage llm-help’ could print out markdown. Thinking about how tools like Claude Code could learn to use new stuff without having to search the web. @jeremyphoward

Of course there are some bad bugs in huggingface’s chat template- https://t.co/JOQYDYAgYW