01
Five layers of telemetry
From the health of the whole system down to the attribution of a single parameter. The network computes these as it trains; switching them off would take extra work rather than save any.
Data goes into a model and stops being something anyone can point at. This is the correction.
Standard architecture
one node
- op
- op
- op
- op
The path from one parameter to the loss runs through everything downstream. It has no fixed length, so per-node attribution cannot be computed — only estimated afterwards, by a separate model, which is what every interpretability tool is.
ORMAS
one node
- LayerNorm
- Linear
- PReLU
- Linear
Four operations, through a shared 4,416-parameter bottleneck. The chain cannot grow with the network, so the contribution of each node is read directly off the backward pass.
Data is the asset in every one of these markets, and most of the best of it is never trained on. Not because it is secret, and not because anyone holding it is being careless. It stays where it is because releasing it means putting it inside a model that afterwards cannot say what it did with it, and no custodian signs that.
Train a model the ordinary way and the process erases itself. What survives is weights, a loss curve, and whatever the team still remembers. Six months on, a regulator asks what changed between the version validated and the version shipped, and every available answer is a reconstruction.
The cause is not carelessness. In a standard network every error signal reaches every parameter on every step, so by the time training ends there is no component left that a question can be put to. The information was never lost. It was never separable in the first place.
ORMAS bounds each component's learning path so the component stays individually addressable throughout the run. The data does not change. The objective does not change. The model that ships is the one that would have shipped anyway. What is added is a record of what happened inside it, and parts that can be pointed at by name.
Naming is the whole thing. A part that has a name can be watched while it trains, protected when it starts to fail, corrected inside a limit declared in advance, tied back to the data that produced it, and taken out when a contract says it has to go. That last property is the one that opens a door. A data owner who can be shown exactly what a model took from their data, and can have it removed on request with a certificate to prove it, is a data owner who can say yes.
WHAT THE ARCHITECTURE GIVES
All of it is produced while the model trains.
None of this is assembled afterwards out of logs. It falls out of the arithmetic the network is already doing, which is why it can be trusted and why it costs nothing extra to generate.
02
Health for every component, at every step
Not sampled and not periodic. Each component is measured continuously against its own baseline, so a component drifting away from itself shows up while it is still drifting.
03
Eight named failure conditions
A component can die, oscillate, saturate, stagnate, explode, lose confidence, lose its gradient, or fail in a way the system declines to classify. Each one has a name and a declared ceiling on the response.
04
Corrections capped before they happen
Nothing is altered by more than a stated fraction of that component's own weight. Every change is balanced so it adds nothing on net, and whatever is left over is written down.
05
A read-only mode
Full diagnostics, zero modification. Most institutions want to watch a system for a year before they let it touch a model, and this is the mode that lets them.
06
A record that can be diffed
Hashed, exportable, and comparable line by line against the version last approved. Reviewers rarely ask what the model is. They ask what changed.
07
Runs where the data already is
No hosted service, no metering, nothing leaving the network. Inside a hospital enclave or a bank, this is what decides whether the conversation happens at all.
Seven features of one product, worked out end to end.
Six operations on a named part, being protect, repair, account, detect, isolate and remove, combine into a large number of useful things. These are not seven products and there is no seven-item price list. They are seven features of one architecture, and each is named after the job it does with data rather than after the mathematics that does it, because the person who signs for it owns the data problem. Only these seven are claimed, each aimed at a sector already paying for its absence, and each backed by evidence available on request. They appear in the order of what each one asks for.
01
The Proofreader
Which samples are damaging which component?
Attaches to an already-trained model without modifying it and returns a ranked list of labels judged incorrect, held separate from those that are merely difficult. Broken down by class, and by source where the corpus carries one.
Read this application →02
The Warning Light
Is this run failing, and where?
Watches every component of a model while it trains and raises an alarm when one crosses its own baseline, naming the component rather than the run. The operator rolls back hundreds of steps instead of thousands.
Read this application →03
The Diary
What did the model do to itself, and when?
Produces a signed, timestamped record of every modification made during training: the component, the diagnosis, the treatment, the step, and the limit it remained inside. Diffable against the last approved version.
Read this application →04
The Separator
Which part of this is the source, not the signal?
Separates what a model learned about the world from what it learned about where the data came from. One model ships and works anywhere, carrying a named removable part for each source and a record of the data that justified each.
Read this application →05
Certified Deletion
What data is in here, and can we remove it?
Removes a named source's contribution and issues a signed certificate of exactly what was removed, with a bounded statement of what changed, and without retraining.
Read this application →06
The Federated Node
Can we train together without pooling the data?
Runs inside an existing federation. Each party's contribution is recorded before the averaging step destroys it, so a round that drops six points can be attributed to a site.
Read this application →07
The Update Engine
What is this model allowed to change about itself?
Retrains a deployed model inside a limit declared and signed beforehand, and produces the evidence that every change stayed inside it. Built to the shape of a predetermined change control plan.
Read this application →
Other combinations are possible and some of them are obvious. This page lists what we will sign a contract against, which is why they are not on it. Adding one costs a configuration change rather than a second licence.
HOW IT IS DELIVEREDHow licensing works
There is no menu, and no per-feature price.
One deployment licence covers the architecture and every feature it can be asked to run. A licensee enables what the deployment requires, and the scope goes into the contract rather than onto a meter. Annual, per institution, running inside the customer environment, beside the data rather than anywhere near ours.
Beyond the software, the licence carries the parts that actually decide a regulated deployment: integration into the customer stack, limits calibrated for the domain, builds pinned to a version anyone can reproduce, support, and indemnity. The research and its experiments are released in full, so the mechanism can be checked by anyone who cares to.
Start with the problem already on the table.
Every engagement starts with one specific problem: a run that died, an audit question that had no answer, a dataset nobody can vouch for.
We reply within a few days. Every message is read by a person.
Pre-book a deployment
One licence, no price list. Terms are agreed per institution and written into the contract, and pre-booking fixes them ahead of general availability.
Useful to bring
- What is being trained, and roughly at what scale
- The obligation driving it: SR 26-2, the EU AI Act, Article 17, a PCCP, an internal policy
- What a failed run currently costs to diagnose