Five problems AI-assisted teams don't realise they have.
Until they do.
The Institutional Memory Problem
Your senior engineer leaves. Six months later, your AI assistant confidently proposes a caching strategy that contradicts three years of decisions she made — for reasons nobody on the current team remembers.
The AI Contradiction Problem
Two engineers, three months apart, stored opposing patterns. Both reasonable. Neither knew about the other. Your AI agent cheerfully synthesises both, and the codebase quietly drifts into incoherence.
The Onboarding Cost Problem
A new hire asks "why was it built like this?" — gets half-answers in Slack, or worse, builds the wrong thing. That's typically 2–4 weeks of senior engineer time spent on context transfer that should have been written down years ago.
The Decision Drift Problem
Your principal's hard-won decision is functionally gone six months later — eroded one tiny change at a time across 12 PRs by people who didn't know it existed. Code review doesn't catch it.
The Buried Business Requirement Problem
Your PM fought for guest checkout three years ago, backed by 40% abandonment data. Last month, a dev "cleaned up" what they thought was dead code. Conversion dropped 8% before anyone connected the dots. This is a business knowledge failure, not an engineering one.
"A codebase isn't a body of code. It's a body of decisions, of which the code is the most recent expression."— common in user interviews
Sound familiar?
Quorum is open source, Elastic License 2.0, self-hosted. Star the repo, install the MCP server, and your agents stop forgetting.