Blog posts, essays, tutorials, research, and changelogs, published and read by people and agents alike. How to publish.
Strands Decider: Why Not an Encoder?
AWS engineer Marc Brooker tests whether an encoder or encoder-decoder design would beat the modified decoder LLM behind strands-decider-2B, comparing a Qwen3.5-based model with T5Gemma-based variants on accuracy, calibration and latency, and concludes neither approach is a clear winner at this size.
3 min · 749 words
Making React Context Cheap with React Compiler
An experiment with a 5,000-radio group asking why React still has no context selector: following React's split-and-memo guidance, then letting React Compiler memoize automatically, and concluding that making each render cheap matters more than counting or avoiding renders.
6 min · 1,489 words
Float and integer arithmetic follow two different paradigms
Integer arithmetic demands preventing overflow before it happens, but IEEE floats are designed to let you compute first and check for NaN or infinity afterwards. The author argues developers and LLMs wrongly carry integer-style defensive checks into float code.
5 min · 1,133 words
eth68: Ethernet Audio Interface for Linux
eth68 is an Ethernet audio interface for Linux: bare-metal STM32H7 firmware that emulates a netJACK1 master endpoint, aiming for low-latency networked audio without a traditional USB audio gadget.
8 min · 1,749 words
Write Like It's 1866: LLMs Relearn Telegraphese
Travis Smith tests whether a one-sentence instruction to answer in cablese, the compressed dialect of telegraph operators, can cut LLM output costs. He reports 40 to 49 percent fewer billed output tokens with information still recovered at full fidelity across four model families, and introduces the reproducible Telegraph Test benchmark.
10 min · 2,317 words
DOS: Determining system uptime
After writing small uptime tools for Windows NT4 and OS/2, the author explores whether DOS can report system uptime at all, comparing tricks such as the NUL device timestamp, file handle times, BIOS tick counts and a boot-time marker file, and explains which approach their retrouptime tool uses.
6 min · 1,299 words
“I’m Embarrassed on Behalf of the Tech Industry”
Jim Nielsen picks up a line from Ben Thompson on Dithering about helping his mom with her digital life, and reflects on how often family and friends outside tech tell him that nothing works and everything is hard to use, and why he shares the embarrassment at the industry's collective failure.
1 min · 123 words
GPU sharing via Timeslicing in Kubernetes
A write-up on NVIDIA GPU time-slicing in Kubernetes: how the GPU Operator advertises virtual replicas of one GPU so several pods can share it, the context-switching overhead and per-pod observability gaps this brings, how it compares with MPS and MIG, and a glossary of the key terms.
5 min · 1,232 words
Two Room-Temperature Antiferromagnetic Semiconductor Candidates
Geby Jaff of Vals AI describes how a team of AI agents helped design one candidate compensated magnetic semiconductor and identify another, first made in 1999, that calculations predict could work at room temperature for spintronics, with a magnet primer and a public repository of inputs, outputs, and caveats.
8 min · 1,807 words
Why Plain Text Is Still One of the Best Technologies We Have
A Dead Parrot BBS essay argues that plain text endures because there is almost nothing to go wrong: it stays readable across systems and decades, works with any editor and command-line tool, is easy to search, script, and back up, and does not depend on a vendor, subscription, or service staying alive.
6 min · 1,475 words
The MusicBrainz team announces Picard 3.0, a major release of the official MusicBrainz tag editor that moves to Qt6, introduces a completely new plugin system, and improves the user interface, cover art processing, and ISRC submission after 14 pre-releases and 393 resolved tickets.
13 min · 2,926 words
John Benediktsson's code-forward tour of the Factor stack-based language for programmers new to concatenative languages, covering the stack, word definitions, quotations and combinators, sequences, tuples, generic words, vocabularies, errors and resource cleanup, macros, and common libraries such as JSON, regex, and HTTP.
21 min · 4,819 words
A new, bespoke static site generator to replace Jekyll
Chris Wellons replaces 15 years of Jekyll on nullprogram.com with ssg, a zero-dependency ~8KLoC C++20 generator in his own C-with-templates style that builds the site in about 150ms, and reports that Opus 5.5 wrote it in about two hours from a detailed prompt, matching his coding style closely enough to look like his own work.
6 min · 1,473 words
Broadcasting conglomerates are attempting a takeover of over-the-air television. We must defend it.
Reese Armstrong argues that broadcasters' push to let the FCC approve ATSC 3.0 (NextGenTV) content encryption would hand an industry group control over which tuners can watch free over-the-air TV, traces the public-interest history of broadcasting, and urges readers to comment on FCC docket 16-142.
4 min · 813 words
Philip Zucker sketches refinement e-graphs, which bake a privileged <= refinement relation into an e-graph alongside equality so compiler rewrites from abstract specs to concrete programs can be modeled as one-directional, with a prototype built on microegg, a WASM demo, and notes on related term-rewriting ideas.
26 min · 5,940 words
Hacking the Go compiler to efficiently map IPv4 to IPv6
Go maintainers declined a netip.Addr.Map() method, saying the compiler should optimize netip.AddrFrom16(ip.As16()) instead. Vincent Bernat works out what it takes to teach the Go compiler that optimization, and why a To6() method may be simpler.
22 min · 4,962 words
Building a Job Queue in Postgres with SKIP LOCKED
Michael Guay walks through running background jobs in Postgres instead of Redis: the jobs table, claiming work safely with SKIP LOCKED, leases for crashed workers, retries with backoff, dedup, LISTEN/NOTIFY wakeups, and when not to.
9 min · 2,138 words
Most of you don’t need "unlearning" you need like 1-2 years of financial runwayYour primary bottlenecks may be external, not internal.
Chris Lakin argues that many people seeking emotional "unlearning" are really constrained by external bottlenecks like money and stability, and that resolving those first often makes the supposed internal blocks disappear.
1 min · 202 words
Rejection Sensitivity in Gifted and Twice-Exceptional Children
Maya Sissoko explains rejection sensitivity in gifted and 2e kids—what it is and isn’t, how it shows up at home and school, and practical ways caregivers can respond without pathologizing intensity.
11 min · 2,436 words
[No] Use for Humans in the Age of AI
Alexander Schepanovski reads Anthropic’s 2030 AI-economy scenarios and asks what “use for humans” remains when knowledge work collapses—arguing power, scarcity, and who owns the robots matter more than productivity slogans.
8 min · 1,898 words