Hacker News. Daily summary. Top 20 stories.
A mathematician alleges OpenAI raced to claim a Navier-Stokes result after learning of his work.
The thread focuses on Buckmaster's allegations that OpenAI rushed to claim a Navier-Stokes result after learning of his team's progress.
OpenAI claims its AI solved the Navier-Stokes Millennium Prize Problem.
Skepticism over cost, originality, and whether this represents understanding or brute force.
Mistral raised €3B at a €21B valuation for sovereign open-weight AI.
Skepticism about Mistral's model quality versus US and Chinese competitors, but support for sovereign AI as a hedge.
LibreOffice 26.8 set download records after declaring its release has no generative AI features.
Skeptics argue the download record reflects steady growth or AI-tool bundling, not the anti-AI stance.
Ex-Anthropic researcher warns both OpenAI and Anthropic are racing recklessly toward superintelligence.
Deep split: some see AI as more dangerous than nuclear weapons; others dismiss fears as hype from IPO-bound labs.
Google DeepMind releases AlphaGenome Atlas, predicting all 9 billion possible single-letter DNA changes.
Skepticism over clinical utility and commercial restrictions, with praise for the resource but doubts about prediction accuracy.
Meta released Muse, a personal AI agent that browses the web and takes actions for you.
Strong distrust of Meta's data practices dominates the thread.
LG smart TVs log data and record audio even when offline or on standby.
Commenters broadly urge disconnecting smart TVs from the internet or isolating them on a separate VLAN.
A GitHub repository offering coding agent skill files for concise, ADHD-friendly output.
Two camps: those who find simple prompts sufficient, and those who see value in the structured skill. A side thread debates ADHD terminology usage.
Terence Tao suggests open math problems are being non-renewably mined by AI.
Two camps: those fearing AI disrupts math tradition, those noting historical competition.
DaVinci Resolve 21.1 adds AI assistant integration and over 100 new tools.
Most commenters welcome the AI assistant for automating tedious editing tasks, seeing it as freeing creativity.
Turning an e-ink reader into an IPP printer so you can Cmd-P to it.
Two camps: those loving the print metaphor and IPP experts offering protocol improvements to save RAM.
Flock Safety's mass surveillance network faces growing backlash and bipartisan opposition.
Thread splits between those citing global precedents (UK, China) and those warning about automated remote access enabling abuse.
89.6% of European companies using a CDN rely on Cloudflare, per CipherCue data.
Bunny.net is the most-cited European alternative, but commenters flag its immaturity and lack of Cloudflare Tunnel.
Satire claims RTO is needed to use AI in person at a corporate dystopia.
Two camps: those loving the Orwellian mayo-slop satire, and those arguing RTO helps workers compete with AI.
Paramount used a fake grassroots group to manufacture support for its merger with Warner Bros.
Astroturfing denounced as typical; few defended mergers on narrow grounds, but most saw this deal as corrupt.
OpenAI launches ChatGPT Images 2.5 with sharper details, faster generation, and a Sketch feature.
Mixed reactions: impressive speed and features, but persistent flaws like distorted fingers and AI slop stigma.
A DHS predictive policing unit analyzes Americans' financial data and has local police pull them over without specific suspicion.
Commenters compare the unit to Minority Report and debate whether KYC enables surveillance.
Open-weight LLMs near frontier ability will soon enable automated mass hacking; we must fix security now.
Most commenters confirm LLMs already find real vulnerabilities; concern centers on slow patch deployment and legacy software.
Copyright law does more harm than good and should be abolished.
Split between those protecting corporate monopolies and those arguing copyright stifles innovation and equality.