Tushhar

All is built on sand, but we must build as if the sand were stone.

[Tushhar @tushhar22]

😲

[Alexander Goslin @xandurglar]

Introducing InfiniteDiffusion, my independent paper accepted to #SIGGRAPH2026!

I have one RTX 3090 Ti. No funding, advisors, or team. By day I'm a new grad SWE at Walmart.

The paper has two main contributions:
- InfiniteDiffusion: a new approach to infinite generation with diffusion models.
- Terrain Diffusion: the world’s first learned procedural terrain generator.

Here’s why this matters, and how they are connected. 🧵

[Egor Cherepanov @hirasava_ui]

🎉 We released MIKASA-Robo-VLA v1.0.0 — a benchmark suite for studying memory in Vision-Language-Action (VLA) policies for tabletop robotic manipulation.

https://t.co/LLN7sCokx2

🧠 The goal is simple:
make memory evaluation in robotic manipulation more systematic.
👇

[Owen @owenology]

Just found out that Berkeley course staff are writing hooks inside course repos so if a student opens an assignment in Claude Code or Cursor the agent will automatically ping the staff 😵‍💫 well played

[yufeng jin @yjin_1118]

How long should it take to go from 🗣️ "Evaluate π0 on LIBERO" to a real-robot rollout?

We're betting on ⏱️ 1.5 hours.

Introducing 🐚 Nautilus — one prompt. Any policy. Any benchmark. Any robot.

An agentic harness for robot learning + @Anthropic Claude Code plugin.
🔗 https://t.co/CAwfBOlQPD

#Robotics #RobotLearning #LLMAgents #ClaudeCode #Harness

[Tom Dörr @tom_doerr]

Fits 10 million documents into 4GB of RAM

https://t.co/ylQzuS7SuN

[LeRobot @LeRobotHF]

We built a bipedal robot for about $2,500.

A real, mostly 3D-printed robot you can build, repair, simulate, train, and control.

Today we’re releasing LeRobot Humanoid: an open robot-learning platform with hardware, runtime, identification tools, and training environments.

Blog post: https://t.co/zu2etb1NZo
Repo: https://t.co/4myLRUtZ3W

[Wall St Engine @wallstengine]

Cloudflare CEO Prince on how AI changes who gets laid off first:

Two weeks ago I laid off more than 20% of my workforce. I didn’t do it because Cloudflare is struggling. We posted record revenue growth, have strong free cash flow and are adding an unprecedented number of customers around the world. I did it because business is changing, and to win the future, Cloudflare needs to change with it.

We haven’t found another example in U.S. business history of a public company growing at more than 30% that laid off more than 20% of its workforce. Yet what we did is likely going to become the norm over the next year. This is a story about artificial intelligence, but executives and commentators are misunderstanding how it will disrupt business and who will be affected.

AI isn’t coming for builders or sellers, but it is coming for measurers. Tireless, independent, efficient and available, AI systems can now measure an organization with a level of objective detail and precision that was previously impossible even for the best employees.

For Cloudflare, internal audit previously picked a handful of business risk areas to scrutinize each quarter. Now we’re moving to a system in which every business risk is audited continuously. We’re closing our books faster. We’re making fewer mistakes and catching the ones we do more reliably. And, as CEO, I’ve never had better tools to measure exactly how the business is performing, including identifying our rising stars.

The vast majority of those we laid off last week were measurers. We cut middle managers across the organization because AI allows us to have more direct reports per manager while still measuring and mentoring our teams effectively.

We consolidated our operations functions into a single group that can support teams across the business, using AI to gain specific expertise when needed. We significantly reduced our marketing team, which, like in most companies, was teeming with measurers. Across our finance team, we found opportunities to consolidate and automate.

We received almost a million applicants for 1,111 paid internships this summer. The interns we hired are extremely qualified and AI-native. They’re all builders or sellers, and we expect that the majority will get full-time offers.

[Lianhui Qin @Lianhuiq]

Scaling embodied AI starts with automating the environments.

Introducing SimWorld Studio: a self-evolving factory for endless interactive 3D environments where agents act, fail, and learn.

With coding-agent + embodied-agent co-evolution, navigation success improves from 50% → 90%.

1/

[Pavi @_pavidhiman]

RL agents need millions of real interactions to learn but for robotics, that's either extremely slow or expensive

World models let agents learn by imagining trajectories and I wanted to understand how this actually works at the compute level

So, I built DreamerV3 from scratch: autograd engine, RSSM, two-hot distributional critic, gaussian actor, fused GRU kernels, all without PyTorch

The agent learns an internal physics model then trains a humanoid walking policy entirely in imagination with ~3h of training on a T4!

The technical writeup covers the full architecture and some failed experiments:
https://t.co/mrEt2X4UEc

[jade @jadechoghari]

At @ScaleAILabs, we’ve been exploring how to get models to accurately caption large-scale robot and human manipulation videos.

More than 1,000 hours of new demonstrations hit our platform daily from factories, homes, and industrial sites and every episode needs precise action level captions: what happened, what object was used, and where it ended up.

Here’s what we’ve found so far 🧵

[metr0x @metrox_eth]

23 min of pick and place evals at 20x.
90% success rate
ACT v3 025k, trained overnight by the rsi loop.
probably nothing.

[Tyler Westenbroek @ty_westenbroek]

Real-world RL is still too brittle and data-hungry for long-horizon, contact-rich tasks.

We introduce Simulation Distillation (SimDist), which turns large-scale simulated experience into reusable world-model priors for rapid real-world adaptation.

By combining online planning with dynamics adaptation, SimDist achieves high success rates on tasks requiring precision, force, and reactivity.

Play with our interactive visualization to see for yourself: https://t.co/qFGNySxdAl

(1/n)

[Tushhar @tushhar22]

Most people lose before they start.

A man opened a burger shop and called it “better McDonald’s.” Another built a small school and called it “the Harvard of his town.” Both borrowed identity instead of creating one.

The something of somewhere becomes nothing of nowhere.

[Tushhar @tushhar22]

Nobody tells you this: beauty fades faster in the eyes than in the world itself. When your eyes are new, every street feels alive, every moment feels magical. Then habit arrives. Wonder dies quietly. Maybe the world didn’t become ordinary. Maybe you just stopped seeing it.