We recruit high-fit teams (fast shipping cadence, multi-chain or chain-adjacent work, recurring audit or simulation spend) through warm intros and cold outreach so learning is not only from friendly circles.

Interviews record pain, spend, workarounds, and adoption intent in structured categories. A stated falsification rule applies: if high-fit interviews fail to confirm recurring pain or spend at the threshold we set, we treat the workflow hypothesis as false for that segment, not as something we market through anyway.

Model keys follow a bring-your-own-key design: request-scoped use where configured, encryption at rest, gateway authentication on API paths, and rate limits on sensitive routes, so custody risk is reduced relative to long-lived server-side key storage.

That is design intent and partial implementation proof, not a claim of completed enterprise security certification. Gap items (rotation, full fail-closed proof everywhere) remain on the hardening register.

We size opportunity using workflow spend, audit, simulation, coordination time, deploy-prep overhead, not TVL as a software budget proxy. Published SAM/SOM-style figures in our materials are assumption-led models for discipline and prioritization, not reported revenue or a claim that the market is already captured.

Founder governance

Execution is founder-led with explicit guardrails: validate one painful workflow before widening platform surface, track qualified usage over vanity metrics, and widen scope only after repeated pain shows up across multiple high-fit teams, per the ownership model described in technical documentation.