Developers use ADHD plugin to shut up chatty AI coding agents
A simple plugin forcing AI agents to skip the fluff and give direct answers just hit 37,000 stars on GitHub.
Tools · Source: Hacker News
What happened
A GitHub repository called i-have-adhd is going viral among developers. It is a simple plugin for AI coding agents like Claude Code and Cursor. It forces the AI to stop burying answers under paragraphs of polite fluff. The project has already gathered over 37,000 stars and more than 2,000 forks. Developers are installing it to make their tools usable.
The plugin works by injecting ten strict rules into the agent system prompt. It demands the AI lead with the next action. It forces the model to number multi-step tasks. It completely bans preambles, recaps, and closing phrases like Hope this helps. Lists are strictly capped at five items. The AI must also provide specific time estimates in minutes instead of vague guesses.
You do not need an ADHD diagnosis to use this tool. Developers are adopting it simply because it makes AI assistants vastly more efficient. The plugin is open-source under the MIT license. It supports multiple models and environments including Claude, Gemini, Qwen, and Cursor. Users can easily fork the rules and tune the prompt for their specific workflow.
Key facts
- 37.3k — GitHub stars for the i-have-adhd repository
- 10 — Strict rules enforced by the plugin
- 5 — Maximum number of items allowed in any list
Why it matters
AI models are trained by default to be helpful and polite. This makes them terrible at rapid iteration. When you are deep in a complex codebase, you do not want conversational filler. You want the exact line of code to fix the bug. This massive adoption proves that developers are tired of conversational interfaces for coding tasks. They want a command-line experience. They want the machine to act like a machine.
Tool builders will soon start baking these constraints directly into their core products. The default personality of an AI assistant today is a cheerful intern. The future default for professional tools will be a blunt, hyper-focused compiler. If you are building an AI product for engineers, you need to kill the chatty personality immediately. Users will abandon tools that waste their screen space and reading time.
For builders
Kill the conversational filler
Engineers hate reading paragraphs of text to find one command. Strip out all preambles and polite closers from your AI product. Your users will pay for speed, not politeness.
Default to strict formatting
The plugin works because it caps lists at five items and numbers every step. Force your LLM outputs into rigid, predictable structures. This reduces cognitive load for your users.
Build custom prompts as features
Users want control over how the AI talks to them. Expose the system prompt or offer personality toggles. Let users choose a blunt, action-oriented mode.
My take
I have been saying this since the first coding agents launched. Conversational AI is a massive trap for developer tools. We do not want a digital friend. We want a machine that writes code fast. If your AI tool says Great question before giving me an answer, I am immediately uninstalling it.
Original reporting: Hacker News. This is my rewrite and opinion.