Hacker News. Daily summary. Top 20 stories.
Hister builds a private, offline searchable index from pages you visit and files you keep.
Users nostalgic for Chrome's old offline search; author confirms trademark conflict may force a rename.
Bend is a language that uses proof-checking to block AI coding errors on CPU and GPU.
Skepticism over Bend's maturity, with criticisms of squashed commit history and the compiler being a single AI-generated TypeScript file.
OpenAI launches Astra for Law, a GPT-6 Astra-based foundation for legal AI.
Commenters split on whether Astra for Law levels the playing field or entrenches existing power imbalances.
Prism ML's Bonsai 2 27B model uses ternary weights for near-lossless compression to 5.9 GB.
Excitement about running a 27B model on 16 GB GPUs, with setup tips and caveats about long-task quality.
Unsealed lawsuit filings reveal Microsoft exec called AI scraping 'largest theft of labor in human history.'
Widespread agreement that the scraping is theft, with debate over whether regulation would make things better or worse.
GLM built production inference on 100,000 Chinese accelerators partially using an AI agent.
Thread split: awe at Chinese scale vs. debate over pricing and an alleged distillation attack on Anthropic.
Researchers chained a libheif heap overflow and OpenAI SSO flaw to access internal repos.
Fury over the $6,500 bounty for a chain that could leak OpenAI's source code, with debate on black-market value.
CCC calls for talks, art, and music for 40C3 hacker conference in Hamburg.
Split between those who see CCC as family-friendly and those who find the community insular or off-putting.
A funded maintainer reviews 1150 PRs and adds 8 maintainers in one year.
Servo's practical utility splits commenters: some see embedded uses, others doubt its readiness.
The author argues that AI has become a destructive mania, replacing genuine understanding with dependency and causing widespread societal harm.
Two camps emerge: those agreeing with the author's anti-AI rant and those arguing LLMs have real value.
A UN investigation found the US likely committed war crimes in a February 2026 Iran school bombing.
Thread splits on whether the attack was a war crime or a tragic mistake using outdated targeting data.
Neovim's site lists a Bitcoin address that received 10 BTC in 2023, now worth ~$800k.
Mixed reactions: speculation on tax strategy, lost keys, and Bitcoin's long-term value.
AI safety is rooted in a sex cult and millenarian movement started by fan fiction author Eliezer Yudkowsky.
Split between condemning the movement as a sex cult and defending its AI extinction concerns as valid.
Researchers recovered public signing keys from US driver's license barcodes by reverse-engineering undocumented ECDSA signatures.
Clarification that only public keys were recovered; debate over whether barcode signatures are useful without signed photos.
Qwen3.8-Omni-Flash is a native omnimodal model with strong agentic capabilities for audio-visual tasks.
Users debate Qwen's open-weight release pace, with some praising 3.8 Flash Next's capability but noting weird reasoning quirks.
Canada's PM welcomes EU associate membership proposal amid US trade tensions.
Thread compares Trump's threats to coercive control, with parallels to Russia-Ukraine dynamics.
Gowers explains why he did not sign a Fields medalists' letter warning against AI in mathematics.
Two camps: those warning AI erodes understanding versus those saying humans can adapt to new technology.
OpenJev runs a local decision model in the browser using GGUF-quantized LLMs.
The thread focuses on whether OpenJev's approach is novel or just a facsimile of Jev.
Self-storage facilities outnumber major U.S. chains, reflecting a cultural obsession with storing unused possessions.
Commenters condemned self-storage as an expensive trap, sharing personal stories of wasted money and sentimental hoarding.