# Accord

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Oluwaseun Oyediran](https://cerebralvalley.ai/u/Sheun)
- **GitHub:** https://github.com/SheunDiran/Projects
- **Demo video:** https://drive.google.com/file/d/1Dp7R1XATi_8vI0yrU9aBeXK1KVvbP7by/view?usp=drivesdk
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5321

ACCORD — EVIDENCE TO AGREEMENT

A request is not an agreement.

Every project begins with a promise. But somewhere between messages, emails, revisions and changing deadlines, the original agreement becomes difficult to trace.

For freelancers, designers, developers and small teams, this creates a costly problem: scope creep, unpaid additional work, deadline conflicts and disputes over decisions nobody can clearly prove.

Accord is an evidence-backed agreement memory and change-impact workspace that transforms fragmented project conversations into traceable decisions.

Instead of treating every new message as an instruction, Accord separates what was requested, what was confirmed and what has actually changed.

THE EXPERIENCE

Imagine a creative studio agrees to deliver a five-page website for ₦450,000 by November 18. Days later, the client requests an additional dashboard, an earlier launch date of November 11 and no budget increase.

Was the change approved? What happens to the original delivery plan? Which commitments are affected?

Accord makes those questions visible before the team proceeds.

WHAT WE BUILT

1. Evidence Library: Preserves original source text and links decisions to supporting quotations.
2. Decision Ledger: Tracks proposed, confirmed, rejected and superseded commitments.
3. Conflict Detection: Flags contradictory recorded terms without silently choosing a winner.
4. Change-Impact Simulator: Shows how requested changes may affect delivery, QA, scope and communication before approval.
5. Human-in-the-Loop Approval: Requires explicit confirmation or rejection of proposed changes.
6. Audit Trail: Preserves a history of decisions and revisions, with JSON import/export for portability.
7. Meta AI Integration Architecture: Implements a server-side pathway for Muse Glimmer 30B to extract candidate commitments with source quotations and human review.

The product includes a responsive interface, interactive case-study experience and guided product walkthrough.

THE TECHNICAL APPROACH

Accord combines structured decision logic with evidence-grounded language-model integration. Its server verifies candidate quotations against original evidence before presenting them for review. Crucially, the application never automatically converts an AI suggestion into a confirmed agreement.

The deployed prototype uses browser-local storage and Vercel serverless functions. Its AI walkthrough is explicitly identified as illustrative because hosted model authorization was not completed successfully.

THE VISION

Accord is not another meeting summarizer. It is a trust layer between conversation and commitment.

Our goal is simple:

Stop losing the agreement behind the conversation.

Live product: https://accord-evidence-hackathon.vercel.app

## More from Meta Global AI Developer Hackathon

- [Moment](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5318)
- [Backlot](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5319)
- [Newton & Carter LLC](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5320)
- [Whatauth](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5322)
- [Videonerd](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5323)
- [Proof Loop](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5324)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5321. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt