Open Source Radar — October 5: Skill stacks for every harness

Share
Open Source Radar — October 5: Skill stacks for every harness

The trending board this weekend is less about new frameworks and more about portable taste: skill packs and workflows built in one agent harness, then ported everywhere else. Four fresh picks, all verified at the source.

gstack (TypeScript, 135,000+ stars) is YC president Garry Tan's Claude Code setup, opened up: 23 opinionated tools that play CEO, designer, engineering manager, release manager and QA. The repo sits at the top of daily trending, and the appeal is obvious — instead of writing your own planning ritual, you borrow someone else's, then let the stack route work through structured checkpoints (an office-hours interrogation of your idea before any code exists) and a verification gate that won't let a session end until the tests pass. It installs on Claude Code fully, with experimental support for Codex, Cursor, OpenCode and half a dozen other harnesses, and its cross-review skill sends your code to a rival model for a second opinion.


pstack-claude (JavaScript, about 900 stars) is the same instinct from the other direction: a port of Cursor engineer Lauren Tan's pstack skill stack to Claude Code, Codex, Pi, OpenCode, Gemini CLI and Prime Agent. Tell it your goal and its router picks a playbook — bug investigation reproduces the failure first, features go through planning, refactors bring in an architecture review — then hands you the fix plus the failing and passing evidence. It tracks upstream rather than forking freely, ships no telemetry, and is the cleanest example yet of "Cursor-only" workflows becoming a lingua franca across agent harnesses.


text-to-cad (Python, about 17,000 stars) gives your agent CAD superpowers: a library of skills for creating STEP models from plain-language requests, producing dimensioned engineering drawings as PDFs, writing robot description files, checking parts for manufacturability, and slicing models into printer-ready G-code. It's back on daily trending as plugin installs land for Codex, Claude Code, Cursor and Grok Build, with a viewer that shows the model inside the chat. If your agent workflow has ever stopped at "here's a nicely formatted OpenSCAD file, good luck," this is the fix — it covers the path from idea to a file a CNC shop will actually accept.


Index-Translate (Python, 632 stars) is bilibili's open multilingual translation model family, released under Apache 2.0 and buzzing on LocalLLaMA this week. It comes as 2B, 9B and a 35B-A3B MoE preview, all built on Qwen3.5 architecture, with GGUF quantizations already up for llama.cpp and MLX so it runs locally. Bilibili leans on translation heavily across languages, so this is a model trained by a company that needed it in production — worth a look if you want translation on your own hardware instead of an API meter.

Worth watching this week.

Are skill stacks becoming the real moat, now that the harnesses themselves are interchangeable? Tell us in the comments.

Read more

Akhetonics says its all-optical CPU reaches a customer in 2026

Akhetonics says its all-optical CPU reaches a customer in 2026

Light-based computing keeps promising more than it delivers — but one Munich startup has just put a date on its bet, and the interview laying it out is doing the rounds on Hacker News this week. Akhetonics says it will deploy its first commercial machine with a major customer by the end of 2026, with several more planned for 2027. The company, founded by Michael Kissner and Leonardo Del Bino, is building a computer where data enters as light, is switched as light, and circulates through memory

The Week in AI — October 5–11, 2026

The Week in AI — October 5–11, 2026

Every big claim this week turned out to rest on fine print more interesting than the headline: revenue only the company reporting it can define, safety tests sandboxed while the product keeps the web, and a Pentagon phase-out nobody would confirm until reporters kept asking. The week's top 5 1. OpenAI's revenue was $20 billion below the numbers everyone quoted — and the gap was definitional. The Financial Times reported Thursday that OpenAI's annualized revenue runs roughly $20 billion unde

Drone strike shuts a third Yandex data center, taking YandexGPT offline

Drone strike shuts a third Yandex data center, taking YandexGPT offline

Russia's largest tech company is learning what the AI era's infrastructure war looks like from the receiving end — three data centers in four days, and with them much of the cloud layer Russian businesses run on. A Ukrainian drone strike knocked out Yandex's data center in Vladimir early Sunday morning, the third of the company's facilities hit since October 8. The site — reported at roughly 50 MW and designed for about 2,880 server racks — stopped operating completely after the attack, Yandex

Agent teams cost up to 5x more, barely score higher

Agent teams cost up to 5x more, barely score higher

The multi-agent hype train hit a benchmark this weekend — and the grid and the trucking regulators had quiet weeks of their own. Vals AI put agent teams head-to-head with single agents on its Vibe Code Bench, and the teams cost between 1.8 and 5.1 times more for almost no extra quality. The evals company ran GPT-6 Sol and Claude Opus 5.5 solo and in teams across 50 apps at two reasoning efforts; out of four comparisons, only one was statistically significant — Sol at medium effort, where the t