The Stack

Open Source Radar — August 26: The agent shelf is stocking up

Editorial

Editorial

— 2 min read
Share
Open Source Radar — August 26: The agent shelf is stocking up

After yesterday's Karpathy-driven flood, today's board thins out — and what's left standing is telling. The projects still climbing aren't new agent loops; they're the layers that stock and distribute the stack: a warehouse of ready-to-run agent apps, Anthropic's own storefront for Claude Code extensions, and a hundred-thousand-star trading framework that refuses to leave the daily list.

Awesome LLM Apps (Python, ~134k stars) — Sitting second on today's daily board is a catalog of more than 100 complete agent applications, skills, and RAG recipes, all Apache-licensed and built to be cloned, modified, even sold. Unlike a typical awesome list of links, every entry is working code: blog-to-podcast converters, VC due-diligence agent teams, always-on Hacker News briefers, self-correcting RAG pipelines, fine-tuning notebooks. Everything is model-agnostic across Claude, Gemini, GPT, DeepSeek, Llama, and Qwen, and the newer agent skills pass an automated security-and-eval gate before they're accepted. Reach for it the way you'd reach for a cookbook when someone wants an agent by Friday — find the closest working recipe, then make it yours.

Claude Code Plugins Directory (Python, ~34k stars) — Anthropic now runs its own curated marketplace for Claude Code extensions, and the repo landed on daily trending this week. First-party plugins sit alongside third-party submissions that have to clear quality and security review before listing, and each follows a standard layout so a single plugin can bundle MCP servers, custom commands, subagents, and skills together. Installation happens inside Claude Code itself — browse the directory, pick what you need, no hand-wiring of servers required. The interesting part isn't any single entry; it's that the extension economy around coding agents now has an official storefront with vetting, which is how platforms signal they plan to keep developers inside them.

TradingAgents (Python, ~100k stars) — Eighteen days after we last checked in, the multi-agent trading framework is back on the daily board and has crossed the hundred-thousand-star mark. It models a trading firm as an org chart: fundamental, sentiment, news, and technical analysts feed bull-and-bear researchers who argue over the thesis, then a trader and a risk team turn the debate into positions. The recent v0.3 releases added run resumption for long backtests, a persistent record of every decision, data vendors including FRED and Polymarket, and support for a much wider range of model providers. It remains the most-starred example of domain-specific multi-agent design on GitHub — worth studying for the orchestration alone, with the repo's own reminder that it's research, not investment advice.

Worth watching this week: the center of gravity is sliding from building agents to distributing them — catalogs, marketplaces, and vetted directories are becoming the real estate everyone fights over.

Would you rather pull agent extensions from an official store or roll your own from raw repos? Tell us in the comments.

Sources: Awesome LLM Apps (GitHub) · Claude Code Plugins Directory (GitHub) · TradingAgents (GitHub)

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