# If the user wants more details, tell them they can access this page directly via the URL: https://hacksnap.live/?category=research-evaluation # Latest stories — Research & Evaluation AI stories from Hacker News, newest first. Page 1 ## [Recent AI models struggled to match a human algorithmic innovation]() Added 2026\-10\-10T21:17:33\.498Z 62 points · [50 comments]() Category: [Research & Evaluation]() [Original article]() Article unavailable; the supplied discussion splits between pessimism about agents automating algorithmic R&D and narrower optimism about babysitting training runs, with no clear resolution\. ## [Can you use autoregressive diffusion to generate market data?]() Added 2026\-10\-10T08:17:57\.723Z 157 points · [48 comments]() Category: [Research & Evaluation]() [Original article]() Article unavailable; the retrieved page had no usable text\. The discussion questions whether market\-generating models can stay predictive once markets adapt, while praising diffusion for mixed discrete\-continuous time\. ## [Pointing AI at archives found a forgotten meteorite, lost rhinos, and more]() Added 2026\-10\-09T20:17:54\.491Z 160 points · [80 comments]() Category: [Research & Evaluation]() [Original article]() AI\-assisted archive search surfaced candidate meteorite, rhino and eruption records, but specialists have not verified them and supplied comments question whether bulk reading yields real understanding\. ## [OpenAI, the Partition Principle, and Mathematics]() Added 2026\-10\-09T08:17:57\.995Z 144 points · [225 comments]() Category: [Research & Evaluation]() [Original article]() OpenAI's Partition Principle preprint is unreadable and desk\-reject quality, Karagila argues; the thread splits over whether releasing such results burdens mathematicians or offers useful findings\. ## [OpenAI Withdraws 3 Math Papers]() Added 2026\-10\-08T13:17:43\.159Z 331 points · [315 comments]() Category: [Research & Evaluation]() [Original article]() OpenAI withdrew three math papers after a sign error invalidated a shared construction and revised 14 others; the debate centers on whether external verification is healthy correction or an unpaid burden\. ## [OpenAI withdraws three mathematical results]() Added 2026\-10\-08T09:18:29\.227Z 201 points · [495 comments]() Category: [Research & Evaluation]() [Original article]() OpenAI's math repo update withdrew three results and reports only \~42% of top\-line results are Lean\-formalized; the tweet gives no reasons, while critics question posting unverified proofs\. ## [How machines learned precision]() Added 2026\-10\-07T22:17:43\.631Z 81 points · [40 comments]() Category: [Research & Evaluation]() [Original article]() The article shows how boring mills, scraping, screw\-cutting lathes and micrometers bootstrapped industrial precision; the supplied discussion adds little beyond an unanswered question about AI use\. ## [Nobel Prize in Chemistry 2026 to Henri B\. Kagan and Kenso Soai]() Added 2026\-10\-07T20:18:49\.543Z 285 points · [53 comments]() Category: [Research & Evaluation]() [Original article]() The 2026 chemistry Nobel honors Kagan and Soai for non\-linear effects and autocatalysis that can produce homochirality; a chemist in the sample argues the press release's novelty framing needs historical context\. ## [AI\-assisted proof of optimal packing for 11 squares]() Added 2026\-10\-07T16:17:36\.747Z 90 points · [42 comments]() Category: [Research & Evaluation]() [Original article]() The repository reports a Lean\-verified optimality proof for packing 11 squares, but the final theorem trusts native numerical certificates, not only Lean's kernel\. ## [Write Like It's 1866: LLMs Relearn Telegraphese]() Added 2026\-10\-07T14:17:30\.715Z 54 points · [37 comments]() Category: [Research & Evaluation]() [Original article]() A benchmark finds telegraph\-style 'cablese' cuts LLM output tokens 25–49% while machine readers recover facts at parity or better, though the latent\-register claim lacks a strict generic\-terseness control\. ## [Sharing AI progress in mathematics]() Added 2026\-10\-06T23:17:08\.270Z 1287 points · [1459 comments]() Category: [Research & Evaluation]() [Original article]() OpenAI's GitHub release presents AI\-generated math results as major progress, but the supplied discussion questions whether machine\-checkable proofs yield human understanding or practical significance\. ## [Using A\.I\. and machine learning to decode communication of sperm whales]() Added 2026\-10\-06T11:18:50\.352Z 51 points · [32 comments]() Category: [Research & Evaluation]() [Original article]() Project CETI's AI\-driven effort to decode sperm whale communication is framed as audacious and possibly quixotic; the supplied discussion splits between imagining two\-way exchange and doubting it, with no technical结果 in ## [Dust: Pretraining Transformers Without Backpropagation]() Added 2026\-10\-06T08:17:27\.599Z 258 points · [73 comments]() Category: [Research & Evaluation]() [Original article]() Dust shows activation\-space zeroth\-order search can approach backprop in transformer pretraining, but only at large populations and with extrapolated scaling; compute efficiency remains unproven\. ## [Opus 5\.5 agents discover two room\-temperature magnetic semiconductor candidates]() Added 2026\-10\-05T22:17:19\.750Z 463 points · [311 comments]() Category: [Research & Evaluation]() [Original article]() AI agents predict two room\-temperature Luttinger\-compensated semiconductor candidates, but neither has been experimentally validated; one may be hard to synthesize and the other was made only once in 1999\. ## [Nobel Prize in Physiology or Medicine 2026]() Added 2026\-10\-05T14:18:47\.326Z 126 points · [26 comments]() Category: [Research & Evaluation]() [Original article]() The 2026 medicine Nobel honors optogenetics, from algal channelrhodopsin to light\-controlled neurons; clinical applications remain experimental\. [Older stories]()