autonomous runs8,421 decisions logged127,934 denied by policy134 envelope violations0 control latency< 1 s operation24 / 7 · unattended provenanceverified · 7f3a9d2 raw datanever leaves your network

Nornec Lab OS · live control loop · run #842

The agent recommends. The policy decides. Physics has the final word.

Watch a real control architecture deny its own agent. A proposed setpoint breaches the hard over-voltage plane, is rejected below the model layer, and is replaced by the largest step physics permits — logged, hashed, reversible.

SAFETY STATE ARMED
+ + + +
Pulse voltage
406.0V
hard limit 420 V · within envelope
H₂ flow8.2 mL/min
decision rate3.7 /s
escalationnone required
provenance7f3a9d2
REJECTED
HARD OVER-VOLTAGE ENVELOPE · 420 V
decision provenance 7f3a9d2 · enforced below agent layer
Stage 01 / 07
Measure
47 channels stream from the bench. Provenance is attached before compute touches a single sample.
Decision log — live
— [INIT] loop initialized
drag to orbit · space · → · c · r
8,421
autonomous runs
campaigns kept alive nights, weekends, holidays
127,934
decisions logged
every setpoint, model version and approval — hashed
134
actions denied by policy
the number a dashboard cannot show you
0
envelope violations
in unattended operation — physics enforced below the agent

What we deploy

One auditable loop.
Seven services underneath.

Not a dashboard, not a copilot. A closed control loop on the instruments you already own — on-prem, physics-gated, with human authority over anything high-consequence.

++++
01
Ingest & validate
Multi-channel ingestion from scopes, spectrometers, supplies, PLCs and detectors. Provenance tagged, stale channels rejected before compute.
++++
02
Physics-aware state
Telemetry fused with instrument models and hard constraints. Dimensional checks, conservation residuals, drift correction — continuously.
++++
03
Hypothesis engine
Testable predictions for the next measurement. Nothing becomes a run plan until it survives dimensional, conservation and known-bound checks.
++++
04
Policy & orchestration
Role-based policies convert hypotheses into run sequences. Routine sweeps proceed; high-consequence actions queue for a human.
++++
05
Closed-loop execution
Campaigns run unattended inside hard envelopes for voltage, pressure, temperature, flow and dose. Unsafe drift triggers rollback.
++++
06
Provenance & learning
Every decision carries a traceable hash linking model, policy, input state and operator. Residuals compound into the next campaign.

How engagement runs

Productized autonomy.
Fixed scopes. No slideware.

01
Autonomy Fit Call
We review your instrument park, experiment type and safety constraints — and tell you plainly where autonomy is ready and where it is not.
30 minutes · no fee
02
48-Hour Lab Autonomy Audit
Instrument topology map, agent-readiness scorecard, safety-envelope review and a 90-day roadmap to closed-loop operation.
from USD 12,500
03
Closed-Loop Pilot
One autonomous loop deployed on your hardware: ingest, state estimation, policy, execution and logging — with human escalation.
USD 75,000 – 150,000
04
Lab OS Deployment
Full multi-loop deployment, model governance, runbooks, training and continuous improvement across the laboratory.
USD 250,000+
++++

“The agent may recommend an action. The policy determines whether it is allowed. Physics determines whether it is possible.”

Enforcement order in every Nornec deployment — model, policy, physics

Deploy your first autonomous experiment

Put your lab on a loop.

Start with a 48-hour Lab Autonomy Audit. We map where your lab is ready to run itself — and where humans must stay in charge.

++++ Start the 48-hour audit → Book a fit call