Fetching from the wire…
Public story · 2026-08-07 · high
The project passed 1,701 stars after gaining 85 in a single day, and it's a heavily modified fork of a tool built by Martin's son.
Why now: swarm-forge's star count jumped 85 in a single day, pushing the repo past 1,701 total.
Robert Martin shipped swarm-forge, a tmux-based coordinator that splits AI coding work into six software-team roles, per the GitHub repo. The project passed 1,701 stars after gaining 85 in a single day.
The project began as a fork of a tool his son Justin built. Martin rewrote it heavily in Clojure.
Each of the six roles, Specifier, Coder, Cleaner, Architect, Hardener, QA, gets its own prompt, git worktree, and message channel to the next stage. The whole setup runs over a single, project-scoped tmux socket recorded in .swarmforge/tmux-socket.
Martin demoed the setup building a Clojure and ClojureScript remake of Missile Command with what he calls a six-pack swarm.
The roles run in a fixed order, Specifier to Coder to Cleaner to Architect to Hardener to QA, always in that sequence, per the repo.
The repo doesn't have a license file yet, so the terms for reusing or forking the code aren't settled.
Each link below shares sources, entities, or timing with this story.
us-vs-them (51 points on Show HN, 41 stars) derives line-level provenance from commit authorship and diff analysis, no watermarks or markup, with intermediate scores like 0.46 meaning human-originated but agent-modified. It deliberately models joining, splitting, and dilution...
Robert C. Martin posted that his strategy is to not read agent-written code, instead surrounding it with unit tests, gherkin tests, QA procedures, quality metrics, mutation testing and coverage gates, arguing the gauntlet gives "very high confidence." Booch counters that metri...
Luu re-ran the widely-shared Alderson result (J at 70 tokens average vs Clojure's 109) and showed it was an artifact of trivial Rosetta Code problems. His replacements: implementing a full Zstd decoder from RFC specs with no tests, and the Pandoc task from ProgramBench scored...
Terminal-based coding agent powered by Qwen 3.5. Ships with Qwen-Agent framework and Qwen3-Coder (code-specialized model). Build agentic applications using a completely open-weight stack. GitHub ---
ProgramBench dropped a benchmark that should make every "AI will replace developers" hot take age badly. The setup: give an agent a compiled executable and documentation, then ask it to architect and implement a complete codebase that reproduces the original program's behavior...
The TypeScript project compiles a repository into an explorable, searchable graph you can ask questions against, with the explicit framing "graphs that teach beat graphs that impress." It targets Claude Code, Codex, Cursor, and Copilot. Directly relevant if your agent currentl...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.