Open Source Radar — August 14: agents get an operating system

Coding agents still own the trending board, but today's freshest signal is the layer around them: two all-in-one agent workspaces, an NVIDIA router that points closed-agent tools at open models, and the leading open-source RAG engine holding its daily spot. Four projects worth your attention.

holaOS (TypeScript, ~6.9k stars) — An all-in-one agent workspace that calls itself "the computer for you and your agent": Claude Code, Codex, or its built-in agent run side by side over the same tools, files, browser, and one shared memory, with 100+ integrations plus MCP, frontier models built in or your own keys. Created in March and already past 6.9k stars, it's the fastest-rising bet on the pattern every agent vendor is circling — the agent needs a home base, not a terminal. You'd use it when your agents keep losing context across tools and you want one local-first workspace they all share.

Macro (Rust, ~2.8k stars) — A from-scratch all-in-one workspace that fuses email, chat, docs, tasks, agents, and CRM into a single system with shared team memory, where everything is @-linked and searchable so an agent can follow a thread from a channel message to a task to an email without switching tools. It's the Slack-Linear-Notion-HubSpot replacement built by a startup that dogfooded it for two years, and it hit today's daily board as the purest agent-native office pitch yet. If your team is held together by glue code between five tools, this is the bet that one graph could replace all of them.

Switchyard (Rust, ~1.4k stars) — NVIDIA's open-source answer to model lock-in: a proxy that sits between your coding agent and any model backend, translating between OpenAI and Anthropic API formats so Claude Code or Codex can run on vLLM, NIM, or Ollama without a rewrite. It also spreads traffic across several models for A/B benchmarking, routes by signal — an LLM classifier or your own algorithm — and exposes Prometheus metrics on everything. Pre-alpha and explicitly not production-ready, but it's the most credible sign yet that NVIDIA wants open models trivially reachable from the closed-agent tools everyone already uses.

RAGFlow (~88k stars) — The leading open-source RAG engine crossed 88k stars and keeps showing up on daily trending, and its pitch has sharpened: it's not just retrieval anymore, it's an agentic context layer — deep document understanding feeding agents grounded, citable context instead of keyword hits. Development is active (v0.26.4 is the current release), which makes it the default answer for teams that want RAG an agent can actually trust. If your agents keep hallucinating on your own documents, this is the reference implementation to measure everything else against.

Worth watching this week: holaOS and Macro are pre-1.0 bets on the same idea — if either sticks, expect the coding-agent vendors to ship home bases of their own.

Would you move your whole workflow into an agent-native workspace, or keep your agents in the terminal? Tell us in the comments.

Sources: holaOS (GitHub) · Macro (GitHub) · Switchyard (GitHub) · RAGFlow (GitHub)

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

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

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

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

Read more

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

The Week in AI — October 5–11, 2026

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

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