Blueprint turns one real site into a decision a robot team can trust.
A robot team usually has one facility and one workflow question before a field visit. Blueprint exists to make that exact site legible earlier — as capture-backed evaluation runs, with rights, privacy, and provenance kept readable the whole way through.
We are not a generic AI marketplace, a model demo, or a deployment guarantee. We package capture truth and frame policy comparison as honest estimates, so the next test is chosen on evidence instead of assumption.
- Episodes / run
- 100–500
- Rank correlation
- 0.929
- Surfaces
- 4
- Proof boundary
- Always on
Per policy evaluation
Sim-to-site fidelity
Site · Task · Scenario · Eval
Review support, not proof
Review support · not real-world proof“The expensive part of robotics is field time. Our job is to make one real site usable before that clock starts — with proof a serious team can actually read.”
Blueprint was built by Nijel Hunt around the gap between an interesting robotics demo and serious, site-specific deployment work. Background in robotics simulation, 3D capture, and deployment operations.
What we hold to
Four principles that keep the product honest.
These are the rules that decide what Blueprint shows, what it labels, and what it refuses to claim.
Capture first, claim later.
Every world model starts from one real place. We package the capture truth — where, when, how, and under what rights — before any evaluation output is shown.
Estimates, never guarantees.
Policy comparison is framed as rank fidelity and predicted success on captured tasks. We do not promise field deployment or guaranteed outcomes.
Generated media is review support.
Simulated and generated frames help a team read a run. They are always labeled as review support — never presented as real-world proof.
Rights stay attached.
Rights, privacy, provenance, and hosted-access boundaries travel with the listing and manifest, not with marketing copy.
Next step
Start with the public proof or bring one exact site.
Browse site packages to evaluate the proof style first, or contact Blueprint when the readiness question is already known.