Held-out Monitors Sometimes Degrade, Even When Not Trained Against
Thanks to Dennis Akar, Rauno Arike, Shubhorup Biswas, Claude Fable, Max Heitmann, Vladimir Ivanov, and Jordan Taylor (alphabetical order) for helpful comments on this draft and…
Communities 7074 stories
Google shuts down Nobel Prize winning AlphaFold
Article URL: https://www.engadget.com/2225849/google-shuts-down-alphafold/ Comments URL: https://news.ycombinator.com/item?id=49098375 Points: 45 # Comments: 6
Those who use many layers in CPU/RAM and some in GPU – what are your specs and speeds?
I am trying to figure out if it's worth upgrading my RAM, but I've noticed that some MoE models don't seem to do well with many…
Show HN: Qwen Scribe – local transcription and dictation for Apple Silicon
Article URL: https://github.com/VladUZH/qwen-scribe Comments URL: https://news.ycombinator.com/item?id=49098260 Points: 8 # Comments: 0
A.I. Companies Are Recruiting Electricians and Carpenters by the Thousands
Article URL: https://www.nytimes.com/2026/07/29/business/economy/data-center-electricians-training.html Comments URL: https://news.ycombinator.com/item?id=49098198 Points: 10 # Comments: 0
Self-hosting Kimi K3: 20% more hardware cost, 20% better task resolution
Article URL: https://aistack.imec-int.com/blog/gpu-self-hosting Comments URL: https://news.ycombinator.com/item?id=49098130 Points: 6 # Comments: 0
Graph Engineering Needs a Compiler
Article URL: https://fluxtion-playground.dev/blog/2026-07-29-graph-engineering-needs-a-compiler Comments URL: https://news.ycombinator.com/item?id=49098035 Points: 5 # Comments: 0
I keep coming back to Qwen… Over and Over. Is there really nothing better under 120B?
I was looking for a strong coding model and a strong general model, both should be 120b or under. After weeks of researching, qwen3.6 27b(general) and…
CipherX applies painless permanent tattoos with dissolving microneedle patches
Article URL: https://www.designboom.com/technology/cipherx-painless-permanent-tattoos-dissolving-microneedle-patches/ Comments URL: https://news.ycombinator.com/item?id=49097908 Points: 11 # Comments: 4
Choose DuckDB rather than SQLite
Article URL: https://tracewayapp.com/blog/sqlite-vs-duckdb Comments URL: https://news.ycombinator.com/item?id=49097730 Points: 57 # Comments: 32
Show HN: A verification browser for AI agents – 13ms windows, one-call checks
Article URL: https://github.com/hongnoul/hwatu Comments URL: https://news.ycombinator.com/item?id=49097727 Points: 6 # Comments: 1
AI in Linux
Article URL: https://drewdevault.com/blog/AI-in-Linux/ Comments URL: https://news.ycombinator.com/item?id=49097649 Points: 6 # Comments: 0
Valve Sponsors Work Bringing Open-Source RADV Driver to Windows
Article URL: https://www.phoronix.com/news/Valve-Sponsors-RADV-Windows Comments URL: https://news.ycombinator.com/item?id=49097591 Points: 25 # Comments: 5
"Uncensored" LLMs are measurably more optimistic than their base models
Hi. Many people think uncensored models are basically the same model that just doesn't refuse, but... I was recently checking whether uncensored models would give me…
The idea: on a CPU the decode speed depends on the active params per token, not the total. My objective is trying to run a 10B at 100tok/s on a mid level PC (No GPU).
On the CPU, batch 1 is memory bandwidth bound. But if token/s = bandwidth / (bytes_per_weight * active_weights_per_token) the total number of parameters doesnt slow down…
Understand Kimi K3 from first principles: a recommended order for anyone trying to understand this beast
Everyone is talking about Kimi K3, but if you jump straight into the technical report, you’ll quickly realize it’s standing on years of research -- just…
TokenTown: A visual way to understand how LLMs work
Article URL: https://laurentiugabriel.github.io/token-town/ Comments URL: https://news.ycombinator.com/item?id=49097003 Points: 35 # Comments: 3
Air Purifiers as White Noise Machines
White noise machines are widely used for sound masking, increasing privacy in offices and helping people sleep. The classic approach is a fan in a box,…
Handbook.md shows that long policy documents do not reliably govern agents
Article URL: https://arxiv.org/abs/2607.25398 Comments URL: https://news.ycombinator.com/item?id=49096969 Points: 38 # Comments: 14
After the AI Crash
Article URL: https://potsandpansbyccg.com/2026/07/29/after-the-ai-crash/ Comments URL: https://news.ycombinator.com/item?id=49096953 Points: 36 # Comments: 31
Show HN: Bullshit Detector – agent skills that fact-check videos and articles
Article URL: https://github.com/SerhiiKorniienko/bullshit-detector Comments URL: https://news.ycombinator.com/item?id=49096917 Points: 26 # Comments: 5
A slide deck you can edit with a local model or in Chrome — the whole deck is a JSON block in one HTML file (~640KB with editor and viewer included)
Over the past few months, our team has been building more and more slidedecks using web frontend technologies with coding harnesses, but a common complaint is…
San Francisco: Don't Fall for Industry Defense of Surveillance Pricing
Article URL: https://www.eff.org/deeplinks/2026/07/san-francisco-dont-fall-industry-defense-surveillance-pricing Comments URL: https://news.ycombinator.com/item?id=49096885 Points: 41 # Comments: 2
Shipping Godot VR and Porting to PSVR2: A Partial Post Mortem
Article URL: https://www.claire-blackshaw.com/blog/2026/07/shipping-godot-vr-and-porting-to-psvr2-a-partial-post-mortem/ Comments URL: https://news.ycombinator.com/item?id=49096811 Points: 4 # Comments: 0