The moat is the dated paper trail, not the prompts
petrova beginner 4 min read
The claim
The moat is not the three files. It is the dated, append-only paper trail every adopting repository accumulates the moment it onboards — and the cross-cutting rule set that file trail enforces — neither of which a wiki, a build tool, or a one-prompt copy can reproduce after the fact.
The evidence
The playbook ships with twelve cross-cutting rules adopted verbatim by every onboarded repository, covering append-only decisions, dated artifacts, and friction checks at every phase boundary. Once a repo is six months in, it has dozens of dated decision notes, a closed-phase trail, and a friction record that any new AI-agent session can read. Importing that trail into a different system is a migration project, not a copy-paste. In the reference workspace, 31 repositories already carry that trail.
Build tools (Nx, Turborepo) cache test runs; they do not capture why a phase closed. Wikis (Notion, Confluence) capture prose; they do not enforce dated, append-only structure that an agent can rely on. Hand-rolled scripts capture neither and rot in months.
The comparable
Linear’s moat is not the issue tracker UI — it is the workflow discipline (cycles, projects, triage rules) that compounds for the customer the longer they run it. Linear reached $400M+ valuation on that thesis (TechCrunch, 2022). Stripe’s moat was never card processing; it was the quality of the API surface and the developer trust that compounded around it. PETROVA’s moat is the same shape: discipline that compounds inside the customer’s repo and travels with them.
What we ask for
Diligence calls with two reference adopters running the playbook for 90+ days. The signal that proves the moat is real: ask them whether they could un-adopt without losing the trail. If the answer is “we would have to rewrite our docs from scratch,” the moat is live.
quadrantChart title Where compounding context actually lives x-axis "Caches build artifacts" --> "Caches project decisions" y-axis "For humans only" --> "For AI agents too" quadrant-1 "PETROVA" quadrant-2 "Internal wiki" quadrant-3 "Hand-rolled scripts" quadrant-4 "Build tools" "Nx": [0.15, 0.25] "Turborepo": [0.2, 0.2] "Notion handbook": [0.7, 0.15] "Hand-rolled scripts": [0.35, 0.3] "PETROVA": [0.85, 0.9]neighbors on the map
- The buyer is the org running ten or more services sizing the addressable market for repo-level AI-agent tooling