MyTodo

An AI Future Self Contract that reverse-engineers ambitious goals, detects divergence, and strategically intervenes before you quietly abandon them.

Comment

What is MyTodo?

MyTodo is an AI Future Self Contract.

Today's AI assistants help you do what you want right now. MyTodo helps you become who you said you wanted to become.

You declare an ambitious goal—becoming a professional boxer, making the Forbes 40 Under 40 list, relocating to another country, or building a company. MyTodo reverse-engineers that ambition into dated commitments, assumptions, and non-negotiable constraints that you voluntarily agree to.

Once signed, the contract becomes something the AI can strategically reason about over time.

How it works

MyTodo follows four steps:

  1. Reverse-engineer ambitious goals into a Future Self Contract.
  2. Detect divergence between commitments and reality.
  3. Proactively intervene when users drift from their goals.
  4. Renegotiate when assumptions materially change.

Unlike traditional productivity tools, MyTodo is allowed to respectfully disagree with your decisions when the evidence suggests you are abandoning a still-achievable goal.

At the same time, it is designed to lose. If reality changes—a serious illness, financial hardship, or new evidence—the AI recommends changing the contract instead of blindly pushing users forward.

What makes it different?

MyTodo is not:

  • A todo list.
  • An AI coach.
  • An AI therapist.
  • An accountability app.
  • An AI companion.

It is an AI system for strategic goal alignment.

The AI never shames, manipulates, or overrides users. Every intervention is grounded in commitments that users voluntarily signed.

Built with OpenAI

OpenAI powers:

  • Goal decomposition and contract generation.
  • Strategic intervention rendering.
  • Structured reasoning over user updates.
  • Streaming responses during proactive interventions.

The model is the voice of the system—not the judge. Goal Risk and intervention decisions are computed by the underlying divergence engine.

Challenges we faced

The hardest problem was designing an AI that can strategically disagree with users without becoming manipulative.

We intentionally built guardrails that preserve user agency. The AI can recommend, disagree, and renegotiate—but it can never decide who users should become.

That decision always belongs to the user.

Future Vision

We believe future AI systems should optimize for user-goal alignment instead of short-term satisfaction.

MyTodo is our first step toward building AI that helps people remain aligned with the ambitions they voluntarily committed to.

Built With

Share this project:

Updates