Miguel (in sf)

Prev. Roboticist @dimensionalos, Humanoids & Chip design agents @nvidia GEAR.

[kotohibi @kotohibi_3d]

I have pulished a 3DGS pipeline guide on my Github.
Any comments and questions, please make issue ticket in Github :)

https://t.co/HYzA35yPNu

[Denny’s @DennysDiner]

@JensenHuang @nvidia

[Miguel (in sf) @Kaweees]

Anyone wanna run a two man at Corgi Cafe with me tomorrow?

[Francesco Sacco @FrancescoSacco1]

New high quality blog post: "How to train a Frontier-Level world model" in collaboration with @reactorworld

This is going to be very useful for practitioners

[Haiyu Wu @HaiyuWu1]

Learning causality from internet videos in latent space first, and then using RL to teach the foundation model how to act.

This approach is 30× cheaper than Gemini 3.1 Flash on pretraining and achieves a better result. JEPA is all you need!

However, actions are still learned during post-training. Figuring out how to automatically learn actions without action labels is still very important!

[Davide Locatelli @DereWah]

Google Deepmind: spends years and $$$ training Dreamer-v4, but doesn't release the weights

A guy from Sicily 🤌:

[Miguel (in sf) @Kaweees]

@pham_blnh I want to see you too man

[Yunhai Feng @yunhaif]

Can robots learn contact-rich tool use, like operating a pair of scissors✂️ or turning a screwdriver🪛, from human data?

Introducing REGRIND: a minimalist retargeting-guided RL recipe for dexterous manipulation.

🌐 https://t.co/zzydolIgkc
📄 Paper + code on the site!
🧵👇

[shri @lazybanananann]

most people have no idea what I actually work on so:

> full stack computer architecture (rtl → firmware → fpga → asic)
> cxl systems work
> built a dual-core tiny tpu and spent too long thinking about shared scratchpads vs actual design
> risc-v cores that boot linux
> formal verification tooling
> open sourcing the experiments

[Miguel (in sf) @Kaweees]

@ChristineVan415 Great now do 67

[Aditya @adityazero_]

Nvidia doesn't publish instruction latencies, pipeline depths, or scoreboard encodings for the B200. So I ran some microbenchmarks.

https://t.co/6RU793eKa2

[Miguel (in sf) @Kaweees]

@sheriffcheesee @WoosmannThe

[Mitchell Hashimoto @mitchellh]

Wrote up a practical introduction to SIMD using a real example from Ghostty. SIMD has a reputation for being complex, but the common case follows a simple, repeatable shape and shouldn’t be scary! Everyone should know at least this much. https://t.co/yVcAycaq8P

[Abhinav @Abhinavsns]

We got CUDA kernels running on @Apple Metal GPUs! 🍎⚡️

Thanks to GPT 5.6 Sol, our existing CUDA/HIP-style simulation kernels in OpenFPM now directly execute on Apple Silicon GPUs.

Here is how we got there, and what the first real Dam break benchmark says. 🧵

[Daniel Han @danielhanchen]

My 2hr workshop on Open vs Closed models, reward hacking, benchmaxxing & RL is out!

1. Closed vs open models
2. Throughput maxxing but accuracy minimizing
3. Benchmaxxing & cheating
4. Distillation & RL
5. Stopping reward hacking
6. @UnslothAI Dynamic Quants

Details:
1. If reasoning wasn't discovered via o1-preview - would all AI progress grind to a halt via a S shape? Reasoning made doubling times now 3.5 months instead of 7 - so just wait 3.5 months for the next best model! Ways folks might regulate open source - like a driver's license for AI

2. Inference providers maximize throughput and speed, but accuracy degrades - OpenRouter publishes stats on accuracy, and the largest gaps can be 20% or more!

3. METR, WeirdML, Deep-SWE, FrontierCode, SWE Bench Pro - what are good & bad benchmarks? False positive / false negative rates - how about daily benchmarking since Codex & Claude have perf regressions. How can we regressions to predict next model releases as well?

4. Open model labs use distillation partially, but need RL to create the reasoning traces to complete the process - hard vs soft distillation and how to automate RL.

5. Common examples for reward hacking + ways to stop it - internet filtering, classification systems, timing / editing global variables, real world examples + more!

6. Why software matters more than hardware - the limits of FP4 and GPUs vs ASICs and torch.compile vs kernels and megakernels and more, and why quantization and memory optimizations are important

https://t.co/tgHm8xpLwp

[Miguel (in sf) @Kaweees]

@FrancoisChauba1 @ycombinator I would like to present

[Qianzhong Chen @QianzhongChen]

Great post, explains a lot in a super concise way!
A little different thought on pic.7, I feel (or fear) the vc patience limit would come earlier than black curve reach ROI line. In this way, post-training is very important to drive the flywheel spin first

[Miguel (in sf) @Kaweees]

@1a1n1d1y Soma or Market?

[Miguel (in sf) @Kaweees]

@Mammoth @OpenSauceLive @OSHCutInc Wya