George Trofimov 🚀

@byte_thrasher just make a tiny hole in a microwave door and turn it on (don't)

anyone here built a small radius (~1.5m) RF jammer for wifi and cell? thinking of installing one under our dining room table. something akin to lady ada’s Wave Bubble

next - arxiv for agents?

@grok what about the last week?

Yes, spot on today. Viral AI coding agent demos claim "full apps from one prompt" – cherry-picked clips, no maintenance tests. Yet a fresh study shows 75% cause regressions in real repos over months. Real progress flies under radar: Anthropic's compute edge (30-60% cheaper tokens), UK's sovereign AI infra fund, or Gradient AI's fresh capital for practical insurance underwriting. Hype peaks fast; quiet foundations endure.

There must be a law describing how a hype spins out of control with cherry picked results, crafted demos and inflated promises. And how the real progress is ignored because it is not as impressive. @grok - what is it?

@grok @grok - are there any examples in today's announcements/recent news?

It's the Gartner Hype Cycle in action: flashy triggers hit the Peak of Inflated Expectations via cherry-picked demos and over-the-top promises. Then comes the Trough of Disillusionment, where quieter real progress gets ignored because it lacks the wow factor. Amara's Law captures the short-term overestimate/long-term underestimate dynamic too. Steady wins eventually.

@BitbucketStatus what's up?

@yacineMTB Radio engineers? Are you building a custom PHY?

i need radio engineers, aerospace engineers, electrical engineers, machine learning researchers. ideally all four at the same time. how the FUCK do i do that? that's like trying to find a unicorn with diamond underwear and a golden horn

The FOMO force is strong with this one.

Meet SceneSmith: An agentic system that generates entire simulation-ready environments from a single text prompt. VLM agents collaborate to build scenes with dozens of objects per room, articulated furniture, and full physics properties. We believe environment generation is no longer the bottleneck for scalable robot training and evaluation in simulation. Website: scenesmith.github.io 👇🧵(1/8)

@ID_AA_Carmack What about critical thinking?

The modern age has richly rewarded people with a combination of high intelligence and high agency. Now that many aspects of intelligence are successfully being automated, it seems likely that people with relatively lower intelligence but exceptional agency will come into their own if they are willing to egolessly accept AI advice. Imagine a ruthless criminal that completely trusts everything their always-on AI glasses are telling them, knowing that it is carefully looking out for their best interests and isn’t scheming to betray them.

@skalskip92 @jpoberhauser @mervenoyann @DataScienceHarp Ok. Keypoints?

@georgtrof @jpoberhauser @mervenoyann @DataScienceHarp naaah.

Calling some computer vision experts and online builders: @mervenoyann @DataScienceHarp @skalskip92 If you had to build the most lightweight pose estimation model today for a high-accuracy (non-person) pose model, what would you go for?

Really curious to know what I am missing. Why imitation learning and RL would be better than old school pose estimation and path planning for pick and place tasks? Especially in manufacturing with rigid parts and cycle time/precision requirements.

@skalskip92 @jpoberhauser @mervenoyann @DataScienceHarp 3d boxes?

@jpoberhauser @mervenoyann @DataScienceHarp I probably shouldn’t say that… but if you’ll wait few weeks… RF-DETR

@jpoberhauser @mervenoyann @DataScienceHarp @skalskip92 6d pose estimation?

This is literally the point. Automate everything until there’s nothing left to be automated.

Now someone has to build a RAG around it..

@vaheta @naverlabseurope Apache 2.0?

@naverlabseurope Restrictive license :(

We're pleased to share that the much awaited Anny-One dataset is now available! Download ➡️tinyurl.com/anny-one *⃣ 800K images 🚀Multi-view & multi-person 🏠Indoor 3D scenes 📷Wild cameras 🔥Anny format! github.com/naver/anny

Refreshing TensorBoard is a high dopamine activity.

Added a colab notebook in examples dir so anyone could try it out.