September 24, 2026

# Nightly Librarian — Newsletter draft

Run: bde438e4-9d42-4c2e-b223-894e4a632c3e
Started: 2026-09-24T06:03:05.856Z
Completed: 2026-09-24T06:04:20.889Z

## Worth attention

- **OpenAI releases GPT-6 Sol and Luna with ~50% lower API prices**
  https://openai.com/index/introducing-gpt-6-sol-and-luna/
  OpenAI shipped GPT-6 Sol and GPT-6 Luna, two tiers below GPT-6 Astra. Reported API pricing is $2/$10 per M tokens for Sol and $0.10/$0.50 for Luna, about half of GPT-5.6, and OpenAI told VentureBeat these are permanent rates; prompt caching also gets higher default hit rates with up to 90% off cached reads. If you run high-volume classification, extraction or triage calls (like nightly-librarian or CalenCall summaries), re-price your cheapest tier against Luna this week.
- **Anthropic releases Claude Opus 5.5 at lower price**
  https://www.anthropic.com/claude-opus-5-5
  Anthropic released Claude Opus 5.5, the first of its 5.5 family, priced at $4/$20 per M tokens with cache reads cut to $0.20/M, about 40% cheaper to run than Opus 5 and 30%+ faster output. Anthropic reports it beats Fable 5.1 on agentic coding, computer use and knowledge work; it is live on the API, AWS, Google Cloud and Azure, with Sonnet 5.5 and Haiku 5.5 promised soon. If your agent/MCP pipelines run on Opus 5, swapping the model string is a near-free cost and latency win worth testing now.
- **tokens too cheap to meter**
  https://jyn.dev/tokens-too-cheap-to-meter/
  An essay arguing the cost of getting a task done with machine intelligence is falling by orders of magnitude per year, and working through what that does to supply and demand. It lands the same week as two big API price cuts, which makes it a useful framing piece. Worth a read if you are deciding whether to design products around token cost as a constraint at all.
- **FoxScript: Visual FoxPro revived on a Rust/wasm runtime (MIT)**
  https://foxscript.org/
  A developer rebuilt the Visual FoxPro 9 language on a new Rust runtime compiled to wasm, checked against the real vfp9.exe, removing the 2 GB table limit, still loading old 32-bit .fll add-ins, and adding lambdas, JSON and an HTTP server. It was built for a paying customer who wanted to keep a 20-year-old business app alive; reports are unfinished and builds unsigned. Mostly interesting as a proof of the "keep legacy business apps running" market for solo builders.
- **Fearless SIMD v1.0 released (Rust)**
  https://linebender.org/blog/fearless-simd-1-0/
  Linebender shipped v1.0 of Fearless SIMD, a Rust library for portable, safe SIMD code. A 1.0 signals API stability for Rust developers who need vectorized hot paths. Only relevant if you write performance-sensitive Rust.
- **GPT-6 Astra reportedly breaks long-unsolved Enigma message**
  https://www.cryptocellar.org/bgac/the-mvueh-break.html
  A cryptography-history site reports that GPT-6 Astra helped break an Enigma message unsolved since 2005. It is an interesting capability data point, but a single anecdote rather than a workflow change. Keep an eye out for how the model was actually used.
- **Hackers claim to have data on all FBI employees**
  https://www.404media.co/we-hacked-the-fbi-hackers-say-they-have-data-on-all-fbi-employees/
  404 Media reports a hacking group claiming to hold data on all FBI employees. The claim is attacker-sourced and not yet confirmed. No direct action for a solo developer; watch for confirmation or details on the entry vector.

## Full digest

- [R] [lobsters] Plain-text files are at risk — https://paste.sr.ht/~awal/b76caf6f213a96a634454fa7292c035a10035b49 — A Lobsters-linked paste arguing that plain-text files are at risk. Content could not be read and there is no identifiable concrete change behind it.
- [R] [lobsters] No Sloptober — https://no-sloptober.com/ — A campaign site urging people to avoid AI-generated "slop" for October. Advocacy, not news.
- [R] [lobsters] That About Wraps It Up for Stock Mac UI — https://inessential.com/2026/09/22/that-about-wraps-it-up-for.html — An opinion post on the state of stock macOS UI conventions. Mostly commentary for Mac app developers.
- [P] [lobsters] Fearless SIMD v1.0 released (Rust) — https://linebender.org/blog/fearless-simd-1-0/ — Linebender shipped v1.0 of Fearless SIMD, a Rust library for portable, safe SIMD code. A 1.0 signals API stability for Rust developers who need vectorized hot paths. Only relevant if you write performance-sensitive Rust.
- [P] [lobsters] tokens too cheap to meter — https://jyn.dev/tokens-too-cheap-to-meter/ — An essay arguing the cost of getting a task done with machine intelligence is falling by orders of magnitude per year, and working through what that does to supply and demand. It lands the same week as two big API price cuts, which makes it a useful framing piece. Worth a read if you are deciding whether to design products around token cost as a constraint at all.
- [R] [lobsters] Jev-powered autocorrection — https://levmiseri.com/nospace — A small web demo of text autocorrection. A toy, not a tool.
- [R] [lobsters] How to talk about "AI" without adding to the anthropomorphization — https://buttondown.com/maiht3k/archive/how-to-talk-about-ai-without-adding-to-the/ — Very good to use for documentation of systems and tasks Comments
- [R] [lobsters] GitHub Actions leaking secrets when Miri output is cached — https://blog.rust-lang.org/2026/09/21/github-actions-leaking-secrets-when-miri-output-is-cached/ — Comments
- [R] [lobsters] Redis is not a map you talk to over TCP — https://blog.verygoodsoftwarenotvirus.dev/posts/2026/09/22/redis-is-not-a-map-you-talk-to-over-tcp/ — Comments
- [R] [lobsters] Re: Suggestions on implementing an efficient instruction set simulator in LuaJIT2 (2011) — https://web.archive.org/web/20180603053407/http://article.gmane.org/gmane.comp.lang.lua.general/75426 — Comments
- [R] [lobsters] Raspberry Pi locks down Pi 5 RAM upgrades in firmware — https://www.jeffgeerling.com/blog/2026/raspberry-pi-ram-lockdown/ — Comments
- [R] [lobsters] Sandboxing with minimal effort — https://yorickpeterse.com/articles/sandboxing-with-minimal-effort/ — Comments
- [R] [lobsters] Parsing JSON Objects without intermediate ASTs — https://arthi-chaud.github.io/posts/json-ir/ — Comments
- [R] [lobsters] Serious editors are a commitment (at least for me) — https://utcc.utoronto.ca/~cks/space/blog/tech/EditorsAreACommitment — Comments
- [R] [lobsters] Why do common Rust packages depend on C code? (2023) — https://langdev.stackexchange.com/questions/3233/why-do-common-rust-packages-depend-on-c-code/3237#3237 — Comments
- [R] [lobsters] Did OpenAI solve the wrong Navier-Stokes problem? — https://www.scientificamerican.com/article/did-openai-solve-the-wrong-navier-stokes-problem/ — Comments
- [R] [lobsters] AI Has No Wisdom and Neither Will You — https://alexn.org/blog/2026/09/22/ai-has-no-wisdom-and-neither-will-you/ — Comments
- [R] [lobsters] EvilVM: Forth shellcode (2019) — https://web.archive.org/web/20250418124519/http://evilvm.ninja/ — Comments
- [P] [hn-top] OpenAI releases GPT-6 Sol and Luna with ~50% lower API prices — https://openai.com/index/introducing-gpt-6-sol-and-luna/ — OpenAI shipped GPT-6 Sol and GPT-6 Luna, two tiers below GPT-6 Astra. Reported API pricing is $2/$10 per M tokens for Sol and $0.10/$0.50 for Luna, about half of GPT-5.6, and OpenAI told VentureBeat these are permanent rates; prompt caching also gets higher default hit rates with up to 90% off cached reads. If you run high-volume classification, extraction or triage calls (like nightly-librarian or CalenCall summaries), re-price your cheapest tier against Luna this week.
- [P] [hn-top] Anthropic releases Claude Opus 5.5 at lower price — https://www.anthropic.com/claude-opus-5-5 — Anthropic released Claude Opus 5.5, the first of its 5.5 family, priced at $4/$20 per M tokens with cache reads cut to $0.20/M, about 40% cheaper to run than Opus 5 and 30%+ faster output. Anthropic reports it beats Fable 5.1 on agentic coding, computer use and knowledge work; it is live on the API, AWS, Google Cloud and Azure, with Sonnet 5.5 and Haiku 5.5 promised soon. If your agent/MCP pipelines run on Opus 5, swapping the model string is a near-free cost and latency win worth testing now.
- [R] [hn-top] Waymo transit rewards — https://waymo.com/blog/2026/09/transit-rewards/ — Waymo announced a transit rewards program. Consumer product news with no relevance to solo software builders.
- [M] [hn-top] GPT-6 Astra reportedly breaks long-unsolved Enigma message — https://www.cryptocellar.org/bgac/the-mvueh-break.html — A cryptography-history site reports that GPT-6 Astra helped break an Enigma message unsolved since 2005. It is an interesting capability data point, but a single anecdote rather than a workflow change. Keep an eye out for how the model was actually used.
- [M] [hn-top] Hackers claim to have data on all FBI employees — https://www.404media.co/we-hacked-the-fbi-hackers-say-they-have-data-on-all-fbi-employees/ — 404 Media reports a hacking group claiming to hold data on all FBI employees. The claim is attacker-sourced and not yet confirmed. No direct action for a solo developer; watch for confirmation or details on the entry vector.
- [P] [hn-top] FoxScript: Visual FoxPro revived on a Rust/wasm runtime (MIT) — https://foxscript.org/ — A developer rebuilt the Visual FoxPro 9 language on a new Rust runtime compiled to wasm, checked against the real vfp9.exe, removing the 2 GB table limit, still loading old 32-bit .fll add-ins, and adding lambdas, JSON and an HTTP server. It was built for a paying customer who wanted to keep a 20-year-old business app alive; reports are unfinished and builds unsigned. Mostly interesting as a proof of the "keep legacy business apps running" market for solo builders.
- [R] [hn-top] ReBarUEFI: Resizable BAR for almost any UEFI system — https://github.com/xCuri0/ReBarUEFI — Scheduled agent omitted this claimed item from the completion payload.
- [R] [hn-top] What California is learning from solar panels built over irrigation canals — https://www.kqed.org/science/2002033/heres-what-california-is-learning-from-solar-panels-built-over-irrigation-canals — Scheduled agent omitted this claimed item from the completion payload.
- [R] [hn-top] SAML: A fractal of bad design — https://blog.trailofbits.com/2026/09/21/saml-a-fractal-of-bad-design/ — Scheduled agent omitted this claimed item from the completion payload.
- [R] [hn-top] How did AMD Ryzen get 50% faster in two years? — https://lemire.me/blog/2026/09/18/how-did-amd-ryzen-get-50-faster-in-two-years/ — Scheduled agent omitted this claimed item from the completion payload.
- [R] [hn-top] WordPress: Unauthenticated path traversal leading to conditional RCE — https://github.com/WordPress/wordpress-develop/security/advisories/GHSA-7hp8-65ch-5whp — Scheduled agent omitted this claimed item from the completion payload.
- [R] [hn-top] Claude Opus 5.5 Intelligence, Performance and Price Analysis (Max) — https://artificialanalysis.ai/models/claude-opus-5-5 — Scheduled agent omitted this claimed item from the completion payload.
- [R] [hn-top] Pentagon says overreliance on AI contributed to missile strike on Iran school — https://www.bloomberg.com/graphics/2026-iran-school-attack/ — https://archive.ph/0V37g
- [R] [hn-top] Unreal Agent — https://unreallabs.ai/blog/unreal-agent/ — https://github.com/unreallabsai/unreal-agent
- [R] [hn-top] The current balance of power in open models — https://www.interconnects.ai/p/the-current-balance-of-power-in-open — Scheduled agent omitted this claimed item from the completion payload.
- [R] [hn-top] People hooked on vapes try a new way to quit: cigarettes — https://www.bloomberg.com/news/articles/2026-09-18/to-quit-vaping-some-are-starting-to-smoke — https://archive.ph/FvOhx
- [R] [hn-top] Grammarly will send unhinged messages to all your users if you try to cancel — https://www.reddit.com/r/sysadmin/comments/1wjdpgx/psa_grammarly_will_send_unhinged_messages_to_all/ — Scheduled agent omitted this claimed item from the completion payload.
- [R] [hn-top] Obscura: VPN that can't log your activity — https://obscura.com/#faq-technical — Scheduled agent omitted this claimed item from the completion payload.
- [R] [hn-top] Show HN: JevBench, a reproducible benchmark for typed decision models — https://benchmarkheaven.com/jev-models — Hi HN! I built JevBench because Jev kicks ass, and the world deserves to know how the serious open source and fake lookalike projects really…
- [R] [hn-top] OpenAI is well positioned to fast-follow Jev — https://arcturus-labs.com/blog/2026/09/21/will-openai-eat-jevs-lunch/ — Scheduled agent omitted this claimed item from the completion payload.
- [R] [hn-top] Side-stepping the Secretary Problem, unwittingly — https://www.evalapply.org/posts/side-step-secretary-problem-hiring/index.html — Scheduled agent omitted this claimed item from the completion payload.
- [R] [hn-top] Markdown in /src — https://htmx.org/essays/markdown-in-src/ — Scheduled agent omitted this claimed item from the completion payload.