ayghri/i-have-adhd: Action-First AI

Overview

ayghri/i-have-adhd treats execution friction as an interface problem: put a known next action first, bound multi-step work, keep state and evidence visible, and preserve safety, detail, uncertainty, and user control. Its canonical policy travels through host-specific adapters, persistence, and tests. A small paired evaluation found better judged actionability, concision, and state reporting in tested conditions, but three trials, same-family judging, an impossible no-tools edit case, and an unsupported-cause regression left the release gate failed.

Learning outcomes

Action-first ordering reduces avoidable startup friction when the next move is known, without replacing safety or requested depth.
Visible state, concrete estimates, and verified outcomes help readers continue multi-turn work without reconstructing its status.
One canonical response policy can travel across runtimes, but activation, persistence, discovery, and prompt overhead remain host-specific.
The recorded evaluation showed a promising actionability signal, yet blockers and overconfident diagnosis left the release gate failed.

Keywords

i-have-adhdactionabilitycontinuityinitiationworking memorysession stateadaptersverificationOpenCodePi/OMP

Chapters

i-have-adhd: Answer-to-Action Design

Why ordering matters more than brevity

9 min · Why ordering matters more than brevity

9 min

i-have-adhd: The Friction Model

From working-memory limits to response tests

10 min · From working-memory limits to response tests

10 min

i-have-adhd: Bounded First Moves

Turning known intent into a verifiable path

9 min · Turning known intent into a verifiable path

9 min

i-have-adhd: Visible Task State

Keeping time, outcomes, and next moves aligned

10 min · Keeping time, outcomes, and next moves aligned

10 min

i-have-adhd: Compactness vs Completeness

A quality test for action-ready responses

9 min · A quality test for action-ready responses

9 min

i-have-adhd: One Policy, Many Runtimes

Separating behavior from activation and persistence

11 min · Separating behavior from activation and persistence

11 min

i-have-adhd: Claude Code and Codex Hooks

Opt-in session injection that fails open

10 min · Opt-in session injection that fails open

10 min

i-have-adhd: Pi/OMP State Machine

Keeping choice, status, and context synchronized

10 min · Keeping choice, status, and context synchronized

10 min

i-have-adhd: OpenCode Adapter Trade-offs

When to inject once or transform each turn

10 min · When to inject once or transform each turn

10 min

i-have-adhd: What Counts as Proof?

Separating tests, scores, and release readiness

11 min · Separating tests, scores, and release readiness

11 min

i-have-adhd: Low-Friction AI Limits

When safety, detail, and control take priority

11 min · When safety, detail, and control take priority

11 min

Preview every chapter · Unlock full text, audio, and video from chapter 4

Plans

Ch.1-3 FreeCh.4-11 Premium

FAQ

What does ayghri/i-have-adhd change in AI coding answers?

It changes response structure so a known next action appears early, multi-step work is bounded, and current state stays visible. It does not improve underlying code or diagnose ADHD.

Does i-have-adhd mean the rules are medically proven?

No. The repository presents them as engineered design assumptions, not clinical validation or a treatment, and using the style does not indicate an ADHD diagnosis.

How does i-have-adhd stay active across coding assistants?

It uses one canonical policy with host-specific commands, hooks, context files, extensions, or prompt transforms. Persistence and recovery depend on the host; installation does not automatically imply activation.

What did the i-have-adhd evaluation show?

In the recorded 2026-08-02 three-trial comparison, the candidate's weighted score was 4.473 versus 4.045 for the baseline, but the release gate still failed. Gains were strongest in actionability and state reporting, while an unsupported cause remained a blocker.