AI news · October 11, 2026
REA gives coding agents tools to reverse-engineer installed software
REA adds a reverse-engineering workflow to a coding agent with npx rea-agents@latest setup; its examples inspect Calculator and the Chrome Dino game, so start with software you own and a reviewable test case.
A Hacker News post from October 10 introduced REA, a tool that gives coding agents a workflow for inspecting software and explaining how it behaves. The project asks the user to run npx rea-agents@latest setup, approve a plan, and restart the agent. Its examples include reproducing Windows Calculator behavior, where 200 plus 10 percent becomes 220, and examining the speed increase in Chrome's Dino game.
The product idea is useful because an agent can first gather observations and write an explanation instead of immediately editing files. That is different from a normal coding assistant, which usually starts from source code and a requested change. The examples are demonstrations, not proof that REA can understand a large or unfamiliar application.
Builders should use a target they own and define a fact that can be checked independently. Keep the first run read-only, compare the explanation with a small test, and save the plan. Only then consider a change, with a human reviewing every file and command the agent wants to use.
What you can do with it
Run npx rea-agents@latest setup in a test project, approve the plan, and ask it to explain one behavior you can verify. Keep the result as notes and tests; do not let the agent change the target until the explanation matches.
Our take
REA points at a real gap in coding tools: understanding behavior is often harder than generating a patch. The approval step is promising, but the value depends on explanations that survive independent tests.
Links Try it
Source: Hacker News — Made With Models writes the brief; the reporting is theirs.