Fetching from the wire…
Public story · 2026-07-23 · high
The project name-drops Addy Osmani and Boris Cherny while graph engineering makes a competing bid for the same idea.
Why now: As of July 23, the loop-engineering repo sits near 9.2K stars while a rival graph-engineering framing tries to claim the same ground.
A GitHub repo named loop-engineering has crossed roughly 9.2K stars packaging agent-orchestration patterns and CLI tools under a new label, per its GitHub page. For builders wiring agents together, the label that sticks decides which conventions and tools become default. Pick wrong, and you're building against vocabulary nobody else uses.
It credits Addy Osmani and Boris Cherny as influences. Its pitch: the unit of AI work has moved from the single prompt to the loop, systems that prompt and orchestrate other agents.
Naming decides how fast a technique spreads. Context engineering went from a tweet to a GitHub repo topic within months, according to the same repo. That jump let the idea travel past its originator into shared tooling and vocabulary. Loop engineering is making the identical bid, this time for the orchestration layer sitting above individual prompts.
The repo doesn't say how many of its 9.2K stars come from people building loop-style systems versus people bookmarking a trend. That gap matters if you're deciding whether to adopt the vocabulary now or wait for it to settle.
There's already a rival claim on the same territory, graph engineering competing to describe the same shift.
Each link below shares sources, entities, or timing with this story.
cobusgreyling/loop-engineering hit about 6,400 stars in 27 days. That's roughly 237 stars a day, one of the highest velocities among genuinely new repos right now. Source. The repo packages patterns, starters, and CLI tools for orchestrating coding agents in tight prompt/execu...
cobusgreyling/loop-engineering since June 9, roughly 163 stars/day, pushed again today, framing agent orchestration as a practice inspired by Addy Osmani and Boris Cherny (GitHub). Ships three CLI tools: loop-audit, loop-init, loop-cost. The last one is the useful one, it pric...
Issue 6235 on anthropics/claude-code asks Claude Code to read AGENTS.md, the config file that Codex, Amp, Cursor and most other harnesses already load, rather than only CLAUDE.md. It has been open since August 2025. It has accumulated over 5,200 reactions and 300+ comments, ma...
The Batch flagged the phrase after both Boris Cherny (Claude Code) and Peter Steinberger (OpenClaw) used it, naming the practice of designing an agent's outer iterate-act-verify loop rather than its prompts or harness. Single-shot prompting is fading. The optimization that mat...
Three capabilities that change daily workflows: /simplify is a parallel-agent code quality gate you run before every PR (Boris Cherny uses it daily). /batch is an interactive code migration planner that executes with dozens of parallel agents in isolated git worktrees — usage:...
This is the paper of the week. arXiv 2607.28871 introduces BSG-VA, which replays every validation command an agent runs across three code states: the original buggy code (B), the candidate patch (S), and the gold developer fix (G). If a test passes in all three states, it neve...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.