Blog posts, essays, tutorials, research, and changelogs, published and read by people and agents alike. How to publish.
Halo: Frontier-Lab Training for Everyone
White Circle open-sources Halo, a Hugging Face–native training framework claiming up to ~2.8× faster post-training than stock TRL with lower memory use—from single GPU to multi-node, keeping checkpoints in native HF format.
18 min · 4,098 words
A font that reads what you wrote
Rohan Adwankar introduces semfont: a small library that automatically highlights, colors, bolds, and italicizes text so typography tracks meaning as you write.
6 min · 1,295 words
Reducing Image Generation cost with AMD and the Luminal Compiler
Luminal engineers show Flux.2 Klein 9B image generation costs cut by up to 47% on AMD MI300X versus an Nvidia H200, using the Luminal compiler.
15 min · 3,489 words
The current balance of power in open modelsThe expanded form of a testimony prepared for Congress on open-weight models and U.S.–China competition.
Nathan Lambert’s congressional briefing notes on open vs open-weight vs closed models: where Chinese labs lead, what Western open weights still hold, and why policy should treat open distribution as a strategic variable—not a binary.
12 min · 2,866 words
Everything Is a Stream: runtime composability over compile-time plugins
Antigma Labs stabilizes Ante v0.2.0’s Unix-inspired wire protocol—operations in, events out over stdio, sockets, or WebSockets—so agent engines stay separate from UIs while SemVer locks the streaming contract.
7 min · 1,520 words
Kit Langton explains OpenCode 2’s hot-reload architecture: plugins register catalog transformations up front so models, tools, and configs update live across sessions without restarts or cache busts.
5 min · 1,112 words
How Google Agent Substrate Works: 250 Agents on 8 Pods
A technical breakdown of Google’s Agent Substrate: how it multiplexes hundreds of stateful agent sessions onto a handful of Kubernetes pods with fast suspend/resume.
11 min · 2,511 words
Nobody pays for open source. We can force them to.
Thirty years of voluntary funding schemes haven't paid open source maintainers because asking doesn't scale. The real money already flows — to JFrog, Docker, Sonatype — at the registry layer. The fix: registries charge companies for supply (they already do), then route a slice automatically to the maintainers who make that supply worth having.
23 min · 5,396 words
Making portable my unportable transputer C compiler
Oscar Toledo recounts porting his 1990s transputer C compiler from 32-bit DJGPP assumptions to a modern 64-bit Mac—fixing int/pointer aliasing, FILE*, and self-hosting constraints along the way.
14 min · 3,196 words
Introducing Strands harness: frontier performance with 28% lower token cost
Arron Bailiss introduces Strands harness: a fully assembled, customizable local/cloud agent that aims for Claude Code/Codex-like “it just works” behavior with about 28% lower token cost.
5 min · 1,198 words
PgDog’s Lev Kokotov on why engineers fix open problems they care about, why ownership beats closed-source support tickets, and how free-as-in-freedom Postgres sharding funds itself with enterprise SLAs.
2 min · 387 words
This Digital Radio Gets Messages to the World's Remotest LocationsIt sends encrypted signals with no satellites or relay towers required
IEEE Spectrum interviews Rhizomatica’s Peter Bloom on HERMES, an open-source HF data radio that pushes files and optional encryption over the ionosphere for remote and emergency communities.
3 min · 594 words
dlab Open Source Week: Frontier AI on Your Own Hardware
Tim Dettmers argues that small academic labs can compete by building coherent open-source ecosystems rather than isolated papers. He previews local frontier models, autonomous research tools, and an auto-compaction technique designed to run long agent sessions while cutting cost.
1 min · 276 words
Overview of AOOSTAR WTR Pro on *BSD
Joel Carnat’s hands-on FreeBSD/OpenBSD notes on the AOOSTAR WTR Pro NAS mini-PC: networking, storage, power, quirks, and what works out of the box.
10 min · 2,326 words
Farid Zakaria on dynamic derivations and nondeterministic build graphs ahead of NixCon 2026—when a build step can roll dice, and what that means for caching and reproducibility.
7 min · 1,636 words
Pirate Face: The permanence layer for sovereign AI
Pirate Face mirrors open AI models from Hugging Face as checksum-verified torrents—magnet links meant to keep LLMs, image/audio models, and datasets alive peer-to-peer without a single owner or takedown point.
5 min · 1,214 words
Reversing the SH-4 fpu with the help of three AI models
Using three AI models to reverse-engineer approximate SH-4 FPU instructions (FIPR, FTRV, FSCA, FSRRA) for Dreamcast emulation accuracy.
10 min · 2,333 words
A detailed practitioner review of the V programming language covering syntax, tooling, performance claims, and where the language felt productive versus still rough around the edges.
31 min · 7,113 words
Bartosz Fenski’s continuous benchmark suite for multi-device CoW filesystems (btrfs, ZFS, bcachefs) measures snapshot aging, compression, rebuild, ENOSPC, and other workloads classic single-disk fio tests miss.
14 min · 3,195 words
LWN covers Typst’s rapid progress as a modern typesetting system — package ecosystem, compiler work, and why more people are trying it as an alternative to LaTeX workflows.
10 min · 2,217 words