- 365Language models pack billions of concepts into 12k dimensions (nickyoder.com)
- 196Show HN: Small Transfers – charge from 0.000001 USD per request for your SaaS (smalltransfers.com)
- 24Cex.C – Comprehensively EXtended C Language (github.com)
- 10Show HN: STT –> LLM –> TTS pipeline in C (github.com)
- 252Read to forget (mo42.bearblog.dev)
- 61IBM Technology Atlas (ibm.com)
- 278Removing newlines in FASTA file increases ZSTD compression ratio by 10x (log.bede.im)
- 171Nicu's test website made with SVG (2007) (svg.nicubunu.ro)
- 330Folks, we have the best π (lcamtuf.substack.com)
- 49GuitarPie: Electric Guitar Fretboard Pie Menus (andreasfender.com)
- 19Show HN: I built a platform for long-form media recs (books, articles, etc.) (rhomeapp.com)
- 104Not all browsers perform revocation checking (revoked-isrgrootx1.letsencrypt.org)
- 79Show HN: I reverse engineered macOS to allow custom Lock Screen wallpapers (cindori.com)
- 29PA-RISC Performance and History (openpa.net)
- 120Imperial Tyranny, Korean Humiliation (english.hani.co.kr)
- 44The Mythical Creatures of London (londonist.com)
- 249Addendum to GPT-5 system card: GPT-5-Codex (openai.com)
- 736Models of European metro stations (stations.albertguillaumes.cat)
- 8Steering Committee Retrospective (haskellforall.com)
- 137Celestia – Real-time 3D visualization of space (celestiaproject.space)
- 35Steam, Discord, Twitch, Reddit to testify before Congress over 'radicalization' (polygon.com)
- 40Will I run Boston 2026? (getfast.ai)
- 571I feel Apple has lost its alignment with me and other long-time customers (morrick.me)
- 56GPT‑5-Codex and upgrades to Codex (simonwillison.net)
- 127I launched a Mac utility; now there are 5 clones on the App Store using my story ()
- 6Electricity through osmosis: Japan opens landmark osmotic power plant (newatlas.com)
- 6"Why Everything Pam Bondi Said About 'Hate Speech' Is Wrong" (reason.com)
- 3Chimps consume equivalent of a beer a day in alcohol from fermented fruit (theguardian.com)
- 59Debian Upgrade Marathon: 3.1 Sarge (wrongthink.link)
- 7DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning (nature.com)