Open Source Radar — September 13: agents learn to hack

Today's trending list skews hard toward security: an autonomous pentest system, a red-team skill library for Claude, and a Rust CLI that solves the unglamorous problem of running five AI agents in the same repo without them stepping on each other.

Worktrunk (Rust, 7,391 stars) — A CLI for git worktree management built specifically for parallel AI agent workflows. Each Claude Code or Codex instance gets its own isolated working directory off the same repo, addressed by branch name, with none of the path-juggling that plain git worktrees demand. It's the tool for anyone who's tried to run five Claude Code or Codex sessions against one repo and watched them collide. Dual MIT/Apache-2.0 licensed, and the project's own September note claims it is now the most popular git worktree manager.

PentAGI (Go, 23,640 stars) — An autonomous penetration testing system that spins up a sandboxed Docker environment, deploys specialized AI agents (researcher, developer, infrastructure), and lets them loose on a target with 20+ built-in security tools including nmap, metasploit, and sqlmap. Agents share a smart memory system backed by a Neo4j knowledge graph, so they learn from previous recon and skip dead-end probes. Supports over ten LLM providers including local models via Ollama, ships with a clean web UI, and is fully self-hosted. MIT-licensed and one of the most complete open-source answers to the question of whether AI agents can run a pentest without burning your server down.

Claude-Red (Python, 3,799 stars) — A library of 78 structured SKILL.md files that turn Claude into an offensive security operator. Each file encodes domain expertise for a specific attack surface — SQL injection, shellcode generation, EDR evasion, ADCS abuse — across 23 categories. Drop the repo into your skills directory and Claude auto-loads the relevant skill based on conversation context, so you only pay the context cost for tools in use. Works across the Claude skills system, Claude Code, and manual Claude.ai projects. Aimed at authorized red-team engagements, bug bounty triage, and CTF preparation.

Worth watching this week.

Do you think AI agents should have unrestricted access to offensive security tools? Tell us in the comments.

Sources: Worktrunk (GitHub) · PentAGI (GitHub) · Claude-Red (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