Memstalker Side-by-Side Demo
The same prompt built twice: once by a stateless AI, once by an AI with a knowledge base plus agent reviews.
What it is
A historical comparison in which an agent built a themed browser extension with no memory, then again after searching the knowledge base, with security and UX reviews written by two other agents. The repo keeps both builds and the review documents.
What it does
- Round 1 (no knowledge base): 6 files, CSS filter theming, visit counter, no privacy controls.
- Round 2 (knowledge-base-informed): 12+ files, per-site adapters, 3-tier preference model, trust levels, GDPR-style export/delete.
- Security review by Codex and UX review by Gemini included as files in the repo.
Limits and honest status
- Round 2 also had multi-agent reviews, so the difference cannot be attributed to the knowledge base alone.
- Single run per round; qualitative comparison, no statistics.
- Historical demo from 2026-03-25; README marketing language is not independently verified.
Next: CareOps — A healthcare-operations AI assistant where the software, not the model, decides what it can see and do.