The knowledge base is the moat, not the model

Draft: Zain's voice pass pending. Assembled from his own meetup write-up (2026-07-03); he reviews before this publishes.

Every designer in the room had access to the same models. Same Claude, same GPT, same Figma AI. And yet the demos that landed weren't the ones with clever prompts. They were the ones where the AI clearly knew things the model alone couldn't.

That's the thesis I keep coming back to: the model is rented; the knowledge base is owned.

What I actually run

My setup is one Git-versioned vault of Markdown: projects, decisions, journals, the lot. Every AI session reads from it cold and picks up where the last one left off. State lives in files, not in chat history.

Three things changed once the vault existed:

1. Sessions stopped starting from zero. The agent reads the state file, not a recap I type from memory.
2. Decisions stopped evaporating. Every "why did we do it this way?" has a ledger entry with the trade-offs at decision time.
3. Agents got triggerable. A shopping-screener that knows my budget rules. A build partner that knows my design taste. Context turns a general model into my tool.

The org-scale question

The personal version works. The frontier is scaling it: what does a team's shared knowledge base look like when every designer's agent can read the org's decisions, patterns, and post-mortems? That's the thing I'm building toward next.

Don't get overwhelmed by the tooling. Start with one folder of Markdown and one honest state file. The moat digs itself from there.