Gates 🦞 retweeted
I feel far from wanting to get older,
Not knowing what is around the corner.
Time comes and goes, slips away,
Melting away our years.
I am waiting to change,
To be again young.
- Excerpt from The Race Against Time by V.G. Lopez, a 64-year-old poet, singer & lifelong student of human behavior
Gates 🦞 retweeted
BREAKING: The war in Iran has not ended like Donald Trump promised it would after the midterms ended.
Gates 🦞 retweeted
Elon Musk was once asked to explain first principles thinking in a way that anyone could understand. Instead of giving the business school version (which he has done many times), he gave an example so simple that it revealed something about how his brain actually works, not how he describes it working.
He said: "If you want to make a cheaper rocket, most people look at what rockets cost today and try to make them 10 percent cheaper. That's reasoning by analogy. You're comparing to what exists. First principles says: what is a rocket actually made of? Aluminum alloys. Titanium. Copper. Carbon fiber. What do those materials cost on the commodity market? About 2 percent of the cost of a finished rocket. So the question becomes: why is there a 98 percent markup between the raw materials and the finished product? The answer is: because that's how it's always been done. First principles says that's not an answer."
The 2 percent number is the one that changed SpaceX. When Elon calculated that the raw materials for a rocket cost 2 percent of the finished rocket, the gap between 2 percent and 100 percent became the engineering problem. Every layer of that gap (labor, manufacturing process, supply chain, contractor markups, design complexity, testing redundancy) was interrogated individually. The layers that were necessary were kept. The layers that existed because "that's how it's always been done" were eliminated.
The elimination produced the Falcon 9. A rocket that costs roughly $2,720 per kilogram to orbit. The Space Shuttle cost approximately $54,500 per kilogram. The reduction is not 10 percent. It is 95 percent. The reduction came not from building a better version of the existing approach but from asking why the existing approach costs what it costs and eliminating everything that doesn't have a physics based answer.
The simplicity of the explanation is the vulnerability. Most CEOs explain their thinking in complex frameworks designed to sound impressive. Elon explained his in terms of aluminum prices and percentage markups. The explanation is so simple that it's almost embarrassing for the audience of MBAs who spent two years learning frameworks that a commodity price check renders unnecessary.
The man who thinks in first principles can explain it to a child because the thinking itself is childlike. Children ask "why" until the adult runs out of answers. Elon asks "why" until the engineer runs out of physics based answers. The moment the answer becomes "because that's how we've always done it," the thing being discussed is a tradition, not a requirement. And traditions are eliminated. Requirements are optimized. The distinction between the two is worth $100 billion in rocket economics.
First principles is not a business methodology. It is the refusal to accept an answer that doesn't reduce to physics. Every child does it naturally. Most adults unlearn it. Elon never unlearned it. The man who builds rockets thinks like a five year old who won't stop asking why. The five year old's question, applied with an engineering degree and a hundred billion dollars, is the most powerful tool in commercial space.
Gates 🦞 retweeted
Hand this to your agent:
repo: github.com/areu01or00/qwen0.…
What it does, closed form, zero gradients:
- reads the activation entering mlp.down_proj (the "key") at every answer-token position, layers 17-21
- target per token = that token's output embedding, scaled by how far the model is from saying it
- per layer: dW = R Kp^T (lam*Kn Kn^T + Kp Kp^T + mu*I)^-1 Kp = answer keys, Kn = keys that must not move (~40 general Qs + your contrast set at 10x), lam=3 - spread bottom-up across 5 layers; each layer delivers its share of what's still missing - closed loop: the edited model answers on its own, finds the first token where it drifts, re-solves with that token pushed harder (boost capped 10x, ≤6 rounds)
- saves a normal HF model folder. 5 of 488 tensors differ, everything else is byte-identical
Knobs: --layer (top edited layer, default ~87% depth), --spread (how many layers, default 5), --lam (higher = more local, weaker write)
Gates 🦞 retweeted
Weight Space Programming: a glimpse into continual learning.
Agent edited Qwen 0.8B's weights to inject facts about my X profile directly into the model: my handle, bio, GitHub, and what i build. It edited 2% of the network (5 of 488 tensors). No training, no gradients.
Result:
Before, it had never heard of me ("I don't have access to private information…").
After, it tells you who i am, my bio and GitHub, answer project questions it was never trained on. Everything else it knew is untouched.
Reproduce : Link to reproduction repo in comments. Point your agent to the repo and teach model your own facts.
Gates 🦞 retweeted
Longevity Grok, a.k.a. your personal longevity superintelligence (PLSI), is an open-source multi-agent workhorse designed to keep you alive & healthy for as long as possible. A collaborative effort from @InSilicoMeds & the Sinclair Lab @harvardmed. Visit longevitygrok.si
Gates 🦞 retweeted
We spent $2m so you don't have to.
And so you can learn how to train MoEs.
Mixture-of-Experts LLMs are what powers the AIs you use everyday, and few people in the world know how to train them at scale (a handful at each top AI lab).
This handbook distills everything we learned in meticulously crafted chapters and figures.
Disclaimer: it may take days to read.
We hope to see 1000x more people able to train their own frontier models all over the world! 🤗
Congrats @Nouamanetazi @lvwerra @Thom_Wolf
Closed-source f*cked again...
team that open-sourced Adobe & Microsoft’s stack just released their Reverse Engineering engine after Claude banned REA.
It's called Craft Anything, and it lets your AI coding agent:
~ reverse-engineer websites, desktop apps, mobile apps and games.
~ study how existing software looks and works.
~ rebuild your own version using your preferred framework or language.
~ compare the recreated version against the original.
~ identify differences and refine the results.
it comes with 284 tools and works as a desktop app, CLI, and MCP server.
find the GitHub link below ↓
long live open source!
Gates 🦞 retweeted
I asked Astra to edit Qwen3.8-27B's weights and make it more token efficient.
For ~$100 in GPU cost, it found the weights for the "done thinking" signal and edited them to be twice as strong.
The result: 40% fewer thinking tokens on average on the pelican on a bicycle test:
Gates 🦞 retweeted
This is literally why we released the Pika H3 Flex weights.
Less than 24 hours after we launched, @_akhaliq built a Turbo LoRA with 2-step draft generation. His release includes 4-step fast and 8-step HD variants, compatible with ComfyUI and Diffusers—and now lives under Pika-Labs on Hugging Face.
We’re so proud to build alongside a community that keeps pushing what these models can do. Can’t wait to see what you make next. 🤗
huggingface.co/Pika-Labs/Pik…
Meet Pika H3 Flex, debuting today as the #1 open weight model on DesignArena’s t2v board.
It’s built for professional iteration, with exportable audio stems, resolutions up to 4k, and faster-than-realtime generation. It can produce a 15-second clip with soundtrack in as little as 7 seconds.
So, yeah. It’s kind of a flex 💅
Gates 🦞 retweeted
80,000 lines of code. 92,000 trees. Real-time ray tracing.
Two weeks of development using Claude Code to build a complete off-road racing game from scratch on a custom C++/Vulkan engine.
Source:
github.com/markaren/threepp
Gates 🦞 retweeted
We @huggingface reproduced this paper!!
It's built with ML Intern, an open-source harness to do AI research, train models, and more
Model: huggingface.co/ML-Intern-lab…
Demo: huggingface.co/spaces/ML-Int…
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
asi will be designed not trained
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
Gates 🦞 retweeted
you can get it here for free: x.ai/bot/yc8yWwlDJLRwr0JuyHR…
Gates 🦞 retweeted
We just open-sourced an entire 3D studio.
Introducing 3D OS by Scenario.
It gives your coding agent 6 departments and 80 skills.
📁 3D OS
┃
┣ 📁 Animation
┃ ┣ 📁 Rigging Any Character
┃ ┗ 📁 Motion From a Video
┃
┣ 📁 Art
┃ ┣ 📁 Previs and Storyboards
┃ ┗ 📁 Characters and Creatures
┃
┣ 📁 Modeling
┃ ┣ 📁 Image to 3D
┃ ┗ 📁 Hard-Surface Weapons and Vehicles
┃
┣ 📁 Look Dev and Lighting
┃ ┣ 📁 PBR Materials and Retexturing
┃ ┗ 📁 Lighting and Rendering
┃
┣ 📁 Worlds
┃ ┣ 📁 Walkable 3D Worlds
┃ ┗ 📁 Skyboxes and Splat Scenes
┃
┣ 📁 Pipeline
┃ ┣ 📁 Retopology and UV Baking
┃ ┗ 📁 Game-Ready Export
┃
┗ 📁 Expert Tools
┣ 📁 Blender, Maya and ZBrush
┗ 📁 Unreal, Unity, Godot and Roblox
From one sketch to a game-ready asset, each department hands off to the next.
And the expert tools drive Blender, Maya, ZBrush, Unreal, Unity, Godot and Roblox right on your machine.
One agent with an entire 3D studio behind it.
Runs on Scenario, inside Claude Code, Codex or Cursor.
Comment “3D” and I’ll send you the GitHub repo.
𝕏 just made it much easier to see exactly which of your posts have had their visibility limited
The latest update to “Under the Hood” now shows the specific posts that received visibility-limiting labels, helping creators understand what's affecting their reach
Here's what's new:
• See exactly which posts were flagged
• Identify the labels affecting their visibility
• Understand how those labels impact recommendations
• View individual affected posts when fewer than 5% of posts in the reporting period received that label
This was a highly requested feature, and 𝕏 has now delivered it
More transparency, more control, and a much clearer picture of what's happening with your posts
Check yours here:
teapaw.org/i/jf/under_the_hood


























