A GitHub repository called i-have-adhd has surged past 4,600 stars in a single day, making it the top-trending open-source project on the platform. It is not a novel algorithm, a new framework, or a machine learning breakthrough. It is a plugin — 10 rules of instruction-prompt engineering — designed to make coding agents output the answer first, skip the preamble, and stop burying useful information under layers of conversational padding.
The project's appeal is immediately obvious to anyone who has used a coding assistant recently. Before i-have-adhd, a typical response to a debugging question might start with 'Great question! Let me think about this,' wander through three tangential observations, and eventually reach the actual fix in paragraph four. After, it leads with the fix: 'Run npm install jsonwebtoken@latest, then edit src/auth.ts:42,' followed by numbered steps and one concrete next action.
The 10 rules enforce a strict output discipline: lead with the next action, number multi-step tasks, end with one concrete next step, suppress tangents, restate state every turn, make wins visible, deliver matter-of-fact errors, cap lists at 5 items, and eliminate preambles, recaps, and closers.
The project is based on techniques from The Adult ADHD Tool Kit by J. Russell Ramsay and Anthony L. Rostain, adapted for how an LLM should respond. It is MIT-licensed and works across Claude Code, Cursor, and other agent platforms.
The viral adoption reflects a broader frustration in the developer community with how AI coding agents present information — and an emerging market for tools that optimize not just what agents generate, but how they communicate it.
