Hyperparameter tuning approach question [R]
I am doing some work with cell type classification, where I have 4.3 million cells and 512 features (condensed embeddings from the encoder of a transformer).…
Full feed
Feed 28,135 posts
Every story across every category, newest first. Each card links to the original publisher; daily-brief posts open as editorial pages.
Locally run assistant on a w-10 board on a local Xiaozhi server
Hello, Locallama! "long-time" member here, from the days when the max context window doubled from 2K to a whooping 4K! Now I feel like I am…
Build your own vulnerability harness
Article URL: https://blog.cloudflare.com/build-your-own-vulnerability-harness/ Comments URL: https://news.ycombinator.com/item?id=48854681 Points: 5 # Comments: 0
Star Just Ate a Planet, and It's Not Done Yet
Article URL: https://www.nytimes.com/2026/07/09/science/space/planetary-engulfment-hungry-star.html Comments URL: https://news.ycombinator.com/item?id=48854638 Points: 6 # Comments: 3
Grok only gets better from here
Grok only gets better from herePaata Ivanisvili: Grok 4.5 just constructed an explicit counterexample to hypercontractivity for the Poisson semigroup (the square root of the Laplace–Beltrami…
Quoting OpenAI
[...] Work on web and mobile runs in the cloud. Work in the desktop app can also use local files and desktop apps with your permission.…
> • 110 m² deployable liquid radiator wait, at what temperatures do they intend to run these chips?! What chips, even? Does this whole project assum…
> • 110 m² deployable liquid radiatorwait, at what temperatures do they intend to run these chips?! What chips, even? Does this whole project assume TeraFab…
The untuned 27B beat the tuned 75B as an agent
I have to admit, a lot of people we're 100% correct to make the suggestion to try this model. I am sorry I ever doubted. The…
RT Command Code: Fable 5 vs Grok 4.5 vs GPT 5.5 We put three top-tier models to build a same game challenge. Used Command Code /design, and the exact …
RT Command CodeFable 5 vs Grok 4.5 vs GPT 5.5We put three top-tier models to build a same game challenge.Used Command Code /design, and the exact…
TL win
TL winSichu Lu: a big reason why economists and economic historians have been very skeptical of AI as not a normal technology is that technology diffusion…
Reading into VLM hallucinations using the Jacobian lens
Reading and editing the visual workspace of a vision-language model.Vision-language models hallucinate: ask one whether some object is in a picture and it will happily say…
Are We Guarding Against Backdoors Or Failing To Notice Them? (Part 1 / 6)
This post serves to argue that backdooring evaluations are prone to failures stemming from triggers never reaching models.In backdooring literature, there is a common workflow. Outputs…
A genealogy of AI safety: how directions are born, and how they die (2005-2026)
What this is and how to read it Ask how AI safety reached this point, and most people give you a neat story. A small group…
RT perigeeaero.shop inc (STRAIGHT): “oh why would you even think this is impressive why would this matter the space shuttle was so much more capable…
RT perigeeaero.shop inc (STRAIGHT)“oh why would you even think this is impressive why would this matter the space shuttle was so much more capable” the time.…
How robust are natural language autoencoders to initialization?
Natural language autoencoders are meant to take in an LLM's activation vector and describe in plain text what the model is thinking. However, its training data…
Focus
Article URL: https://boz.com/articles/focus Comments URL: https://news.ycombinator.com/item?id=48854363 Points: 9 # Comments: 0
Late mover advantage Ironically, China has its own first mover disadvantages. They've built their dystopian surveillance system on pretty primitive te…
Late mover advantageIronically, China has its own first mover disadvantages. They've built their dystopian surveillance system on pretty primitive tech, Palantircore-flavored Panopticon handily outperforms itObject Zero:…
whoever does AEO for @resend needs to get a raise, all the leading frontier models keep trying to use resend for emails even when i have existing tran…
whoever does AEO for @resend needs to get a raise, all the leading frontier models keep trying to use resend for emails even when i have…
Toward A Public Science of Model Behavior
This is a linkpost for our essay "Toward A Public Science of Model Behavior" on transluce.org. The full text is reproduced below.Today’s AI systems frequently behave…
GPT-5.6 Terra and Sol are now available in Perplexity and Perplexity Computer.
GPT-5.6 Terra and Sol are now available in Perplexity and Perplexity Computer.
RT Tesla Owners Silicon Valley: GROK 4.5 LEADS ON REAL PROFESSIONAL WORK BENCHMARK New data from Snorkel shows Grok 4.5 outperforming other frontier m…
RT Tesla Owners Silicon ValleyGROK 4.5 LEADS ON REAL PROFESSIONAL WORK BENCHMARKNew data from Snorkel shows Grok 4.5 outperforming other frontier models on real-world professional tasks.On…
Study: "Mommy, do you love your phone more than me?"
Article URL: https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2026.1766665/full Comments URL: https://news.ycombinator.com/item?id=48854247 Points: 23 # Comments: 1
Show HN: I built a free app for New Yorkers to save money on groceries
I built this because I see that grocery savings are achievable in NYC. People usually just go to the store they're used to going to, and…
OpenAI says GPT 5.6 is the ‘preferred model’ for Microsoft Copilot amid breakup chatter
OpenAI's new family of models will continue to power Microsoft's suite of workplace and productivity apps.