▸ Toggle between the thesis and the anti-thesis. Both URLs are shareable.
/thesis · /thesis#anti
The 97% economy has never had a system of record. We're building one — operator-first, address-level, cryptographically attested — and the same substrate that closes urgent ops loops becomes the underwriting fabric that capital plugs into. Six atoms.
For Allometry to be wrong, six things would have to be true. We hold them in mind, sharply, and ship against them. Honest founders publish their own bear case. Six anti-atoms.
Software is 2–3% of every dollar in the field service economy. The remaining 97% — materials, labor, machinery, capital — runs on spreadsheets and tribal knowledge. CRM owns the customer; ERP owns the ledger; FSM owns the dispatch; none of them own the unit of commercial decision. We're building the system of record for what was never modeled.
Field service is conservative. The wedge stalls at design partners; the cohort never crosses 25 operators; Pulse never gets the data density to compound. The 97% stays unmodeled because the buyer doesn't show up — not because the substrate doesn't exist.
Every commercial decision in physical revenue resolves at an address — a site, a building, a parcel. Build the graph there and revenue, capacity, margin, and risk all stack on the same gravitational point. The customer is downstream of the address; everything else stitches.
Underwriting-grade evidence requires structure we can't capture at write-time without breaking the operator's workflow. The "byproduct" thesis fails: operators produce ops data, not underwriting data, and the gap can't be closed without a parallel data pipeline lenders won't fund.
Sixteen modules close urgent loops — margin leak, dispatch chaos, churn, renewal slippage. The signed, hash-chained evidence accumulates underneath, without operators having to think about it. The vault is the asset; SaaS is the vehicle.
ServiceTitan, Salesforce, or NetSuite ship the same primitive — slower and worse, but already inside the operator. The two-sided network never tips our way; distribution beats architecture, and we lose the wedge before the vault matters.
Lenders unlock by vintage: 6mo working capital, 12mo infra debt, 24mo insurance-linked, 36mo rated paper. Vaults are time-locked. Competitors can copy the endpoint shape in an afternoon; they cannot copy native attested data accumulating month by month. A late entrant starts the clock at zero.
Sub-100ms attestation on real inference workloads stays out of reach for the model sizes that matter. The cryptographic moat stays in pilot, lenders lose patience, and "attested vintage" becomes a marketing line — not a primitive lenders can actually consume.
Per-seat SaaS funds the wedge. Per-decision usage tolls the orchestration. BPS on capital flow is the endgame. Each layer pays for the next; each is a real business if the layer above never ships. The recursion is the moat.
Balance-sheet capital's hunger for novel underwriting signals was a 2023–26 window. Rate cuts return, traditional collateral suffices, and the demand side dries up before the supply side scales. Layer 3 stalls; the recursion never starts; we're a SaaS company forever.
Same arc as card networks and clouds: a thin standardized primitive at the bottom; every counterparty plugs in on top. Lenders, insurers, agents, downstream systems all read from the same signed ledger. The category isn't "RevOps" or "FSM" — it's the substrate underneath both.
Block, Stripe, or Shopify Capital extends an attribution-shaped layer into trade and field service before our time-in-vault matters. We become an acquisition target, not a category — the Tola → Block pattern, with us in Tola's seat.
If this scans, send both URLs. The investor who reads only one is the investor who hasn't read either. Pre-seed conviction lives in the gap between belief and refutation.