The Problem
AI coding agents like Cursor and Claude operate with zero institutional memory. They confidently reproduce architectural mistakes your team has already reverted and learned from—sometimes multiple times. Your postmortems, ADRs, and hard-won lessons are invisible to the agent, so it ships operationally dangerous code that passes review until it breaks something you've already broken.
What We're Building
Déjà Vu builds a temporal knowledge graph from your git history, reverts, PR discussions, and postmortems. Before any agent executes a plan, it surfaces similar historically-regressed patterns with links to the original incident. It risk-scores changes in real time, visualizes blast radius across your codebase, and flags spec deviations before code hits production.
Who It's For
Engineering managers and tech leads at growth-stage startups and mid-market SaaS companies (20–150 engineers) already using Cursor or Claude daily, who've experienced a production incident caused by repeating an architectural mistake.
Key Features (Planned)
- Déjà Vu alerts: Detects when proposed changes match historically reverted patterns and surfaces the original postmortem
- Blast Radius Heatmap: Risk-scores every change and visualizes impact across files, modules, and dependent systems before approval
- Temporal Knowledge Graph: Automatically ingests git history, reverts, PR discussions, and ADRs to build institutional memory
We're validating this idea before writing a single line of code. If this resonates with you, we'd love your feedback:
How many times in the last 6 months has your team shipped code that caused the same kind of production failure you'd already fixed before?
Check out the concept page and let us know what you think.
Built by Jenavus — AI-powered business intelligence
Top comments (0)