LUIS ROSALPRODUCT EVIDENCE
← All workINTERACTIVE DEMOOperational Memory

OPERATIONAL MEMORY · AI COORDINATION

Decisions with sources.
Recommendations with owners.

This demo turns synthetic GitHub, meeting, and team-message events into a cited recommendation. The AI can organize evidence; a human still owns the decision.

ROLE
AI systems, backend & product engineering
FOCUS
Provenance, privacy, approval, handoff
DATA
Synthetic source events
LIVE DEMO

Filter the source events, inspect the evidence behind the recommendation, then approve it or hold it for more information.

Operational MemorySafe pilot · synthetic workspace
4 sources connectedEvidence current
SOURCE EVENT · gh-428BLOCKER

Webhook retries can create duplicate actions

PR #428 adds idempotency keys, but the migration path for existing deliveries is unresolved.

✦
SOURCED RECOMMENDATIONRun a bounded launch-readiness gate
Awaiting owner

Keep the pilot limited to one workspace until duplicate-delivery recovery, the rollback rehearsal, and the retention decision each have a named owner and recorded evidence.

↳ GitHub · PR #428↳ Meeting · Launch review↳ Discord · #operations↳ Runbook · Data retention
PROPOSED OWNERPlatform lead
DUEBefore pilot expansion

42-SECOND WALKTHROUGH

The complete coordination loop.

THE PRODUCT PROBLEM

Project knowledge disappears inside tools and conversations.

A useful memory system must retain provenance, detect duplicates, respect permissions, and distinguish a sourced fact from an AI recommendation.

WHAT I BUILT
  • Event-sourced ingestion and normalization
  • Source-citable decisions, blockers, and actions
  • Privacy controls and permission gates
  • Human approval and operational handoffs
THE DECISION PRINCIPLE

AI may recommend the next action. It should also show the evidence and the person responsible for accepting the risk.