Skip to main content
AI & Technical Systems·A local system for choosing the next action

ADHD Execution Agent

The agent is designed around one operational question: what is the smallest useful action to do next given current priorities and available time?

Status

Public local tool

Capabilities

Workflow design · Scheduling logic · Deterministic controls

The problem

A backlog can add friction when the real need is to choose one action that fits the current window, preserve context, and recover quickly after interruption.

What I worked on

The system combines a task source of truth, Google and Apple calendar context, local scheduling logic, recovery commands, SQLite operational state, and an iMessage interface.

Decisions and tradeoffs

Keep priority logic deterministic

Deadlines, scheduling, completion, quiet hours, and priority rules are handled in code. The optional model can rewrite wording but cannot decide or mutate state.

Design for interruption

Commands such as STUCK and RESCUE shrink the immediate workload while preserving the underlying task state.

Keep the tool local

The system runs on the user’s Mac and does not require hosted infrastructure for its core workflow.

Evidence and scope

This is a single-user local tool. It is not presented as a multi-user productivity product or evidence of measured productivity gains.

What this shows

Product design around real constraints, deterministic control of important decisions, and integration of several everyday systems into one execution loop.

Implementation details

Key subsystems

  • Task and calendar integration
  • Deterministic prioritization and rescue modes
  • iMessage interface through an isolated adapter
  • SQLite operational state