Trust is earned in a specific order.
Engineering rigor first. Structured human validation next. Clinical evidence last. This page is our public account of that order: what has been verified, what is in preparation, and what we refuse to claim early.
Three stages.
No skipping ahead.
Stage 1
Engineering rigor
Built and systematically verified
Deterministic behavior, frozen benchmarks, measured calibration, and end-to-end verification of the platform itself, the level of proof software can provide.
Stage 2
Structured human validation
Designed · recruiting reviewers
Blinded, contamination-controlled expert review with real authority to reject. The framework is built and tested; the human study has not yet run.
Stage 3
Clinical evidence
Not yet begun
Prospective evaluation in clinical settings under the appropriate regulatory pathway. We will not describe the platform as clinically validated until this exists.
Validation as an engineering
discipline.
These practices run continuously, on every meaningful change, not once, before a launch.
Deterministic replay
Every analysis can be re-run bit-for-bit. When something is wrong, we can find it; when something is right, we can prove it wasn't luck.
Frozen benchmarks
Capabilities are measured against fixed, versioned benchmarks, and the failure cases are part of the record, not an embarrassment to hide.
Measured calibration
We test whether confidence scores mean what they say, and correct them when they do not. Overconfident AI is a safety defect.
Adversarial evaluation
We actively look for the conditions that break the system, whether unfamiliar procedures, degraded video, or distribution shift, and log what we find.
Blinded human review
The expert validation program is designed like an experiment: pre-specified analysis, contamination controls, and reviewers empowered to say no.
Versioned governance
Validation decisions are recorded, versioned, and non-deletable. What was accepted, by whom, on what evidence: always answerable.
What we claim today,
and what we don’t.
If a sentence on this website ever disagrees with this ledger, the ledger wins.
Built and systematically verified
- A working end-to-end platform, from surgical video to structured output
- In-house perception models, measured on defined benchmarks, including their failure modes
- Deterministic replay of full analyses
- Confidence calibration, measured and corrected
- A human-validation framework with automated contamination controls
In preparation
- Expert human validation with independent surgeons and reviewers
- Evaluation on broader procedure coverage and unfamiliar conditions
- Hospital pilot programs under clinical governance
Not yet, and not claimed
- Clinical validation of any kind
- Regulatory clearance or approval
- Use in live surgical care
- Claims about patient outcomes
Our commitment: the words on this site track the evidence, not the ambition. Trigone is investigational. It is not a medical device, it is not clinically validated, and it does not make decisions for surgeons, and until each of those changes through the proper process, you will keep reading it here.
Rigor is better judged up close.
Clinical partners see the full validation record: protocols, results, and failure modes included.