While skill optimization for autonomous agents has gained traction, existing methods rely on complex pipelines. This leaves a fundamental question unaddressed: What constitutes a minimal viable…
This update brings Codex as a new agent provider in public preview, expands the Customizations editor with Hooks support and richer MCP server management, and introduces…
Game worlds have traditionally been built through labor-intensive production pipelines, making them costly to develop, difficult to customization, and expensive to modify after deployment. Recent advances…
server: enforce prompt cache RAM limit (#25070) Before this commit, --cache-ram was not a hard limit: The cache always kept at least one entry, even if…
We introduce Nemotron-Labs-Diffusion, a tri-mode language model (LM) that unifies AR, diffusion, and self-speculation decoding within a single architecture. Trained with a joint AR-diffusion objective, Nemotron-Labs-Diffusion…
Scaling robot learning requires massive, diverse trajectory data, yet collection is currently bottlenecked by physical teleoperation, where every demonstration binds operator time to specific hardware and…
We’re really interested in how we ended up living in a democracy - and we think that you should be too. Not only is it just…
Preface: DO NOT READ IF ACTIVELY DEPRESSED OR SU*CIDAL. i do not endorse such actions or views, i merely speculate in order to then find honest…
Polysemanticity is one of the most important concepts when it comes to mechanistic interpretability, basically, studying the internal representations of neural networks. For some reason, the…
By Sohybe Ibrahim Abdelwahab Amer | June 2026The Google DeepMind mechanistic interpretability team (Neel Nanda et al.) suggested a deliberate shift; instead of relying on reverse-engineering…
it_happening.gifTaylor Blau: First day at @OpenAI today!Couldn’t be more excited to start working on SCM tools. Many thanks to @tnm, @gdb, @aglover, @iamveeve, @udayruddarraju, and lots…
common : add missing include in common.h (#25220) Signed-off-by: zhangrunda zhangrunda1234@outlook.com macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED macOS Intel (x64)…
a quiet day lets us read some condensed insight
Article URL: https://www.economist.com/interactive/finance-and-economics/2026/07/02/is-the-economist-always-wrong Comments URL: https://news.ycombinator.com/item?id=48826703 Points: 13 # Comments: 3
I've tuned out Russian-Ukrainian news. It's depressing every time I do look. Russians are starting to REALLY feel the war. My relative failed to fill the…
Hey guys, Despite browsing this sub, and the myriad of other LLM reddits, I feel no closer to understanding what the best memory system is these…
Robotic manipulation in the open world requires not only recognizing what a scene looks like, but also anticipating how its 3D structure moves under interaction. We…
Agentic video understanding equips models with long-term memory to autonomously process and respond to continuous, long-horizon multimodal streams. However, advanced video agents often rely on ``detective-style''…
Xiaomi can legit build Company Towns (within cities). They make next to everything. It will be cheap and good. Infinite IKEA but for real.The only question:…
ggml-hip : add -fno-finite-math-only alongside -ffast-math (#25373) -ffast-math implies -ffinite-math-only under ROCm/clang 22, which disables INFINITY/NaN and triggers -Wnan-infinity-disabled (errors under -Werror in CI). Re-enable infinity…
Article URL: https://stats.lineageos.org Comments URL: https://news.ycombinator.com/item?id=48826329 Points: 11 # Comments: 1
Lots of interesting details in Nemotron-Ultra reportpost-training is a very quickly evolving proto-scienceλux: some interesting post-training details from the nemotron-3-ultra tech report:1/ i like how MOPD…
RT goldenlabubuwatchWatch for Trump deciding to negotiate memory prices himself and cut these numbers in half.Hynix making +usd500bn profit, more than Microsoft plus Google plus Amazon…
Codex for creating the first video of a black hole:OpenAI Newsroom: You may remember the first image of a black hole that computational astrophysicist Chi-kwan Chan…