The irritation
Arguments often continue because each person is responding to a different version of the dispute and nobody wants to volunteer as the neutral judge.
ExperimentPrototypePrototype
SettleIt AI
A low-stakes argument settler that gathers both sides, asks an AI judge for a balanced ruling, and emails both people.
- Current state
- Prototype — Prototype
- Jack’s role
- Concept, Full-stack engineering, AI workflow design
- Find it in the workshop
- Prototype Bench · Argument scales
Built because…
The real build story.
What I built
I built a multi-step application where the initiator can enter both perspectives or invite the other person through a secure email link. Once both sides arrive, an AI model produces one ruling and the system emails it to both people.
The hard part
The application has to collect enough context to compare two accounts without presenting probabilistic model output as legal, therapeutic, or objectively authoritative judgment.
Under the casing
Proof, not decoration.
Build evidence02 items
- A two-party submission and emailed-ruling workflow
- PostgreSQL persistence with validation, rate limiting, and security headers
- Express
- PostgreSQL
- OpenAI
- SendGrid
- Rate limiting
- Input validation
Scope of work
What I owned.
- Concept
- Full-stack engineering
- AI workflow design
No victory-lap math
Outcome and lesson.
Where it landed
SettleIt AI remains a working prototype for playful, low-stakes disagreements rather than a professional dispute-resolution service.
What stayed
AI can provide structure and a fresh framing, but the product must be honest about where generated judgment stops being appropriate.
This project has a physical address.
In the playable workshop, SettleIt AI lives in the Prototype Bench as a argument scales. You can inspect it without unlocking anything.