The number Marta Vieiro gives is 4,112, and she gives it the way people give the year they were born, without looking it up. That is how many firmware releases she certified for Anatome between 2042 and this spring, each one authorising a change to the operating envelope of somewhere between eleven and nine hundred thousand limbs. She read none of them in any sense a tribunal would recognise, and she is the first person to say so.
Vieiro is a Conductor, Prosthetics, Class II, which is worth unpacking for anyone who has never had cause to look closely at the title. The review is performed by the certifying model, which reads the release, tests it against the safety corpus and returns a summary with a recommendation. What the model cannot do is carry the liability, since no insurer will write against a recommendation and no tribunal will accept one as consent, so the last sixth of the work belongs to a person whose signature is the thing that makes the release lawful. Her queue allotted her forty seconds an item. The releases averaged a little over nine hundred pages.
I want to be careful about the conclusion I draw from that, because the obvious one is slightly wrong. Forty seconds does not describe an individual failing on Vieiro’s part so much as the designed throughput of an arrangement that needs a human name at the foot of a document and has priced that name accordingly. She was not employed to read. She was employed to be accountable.
There is a further difficulty, which she raised with her supervisor twice and which I have since put to three engineers outside the company. What arrives in a Conductor’s queue is a summary of a differential taken against a compiled image. The source is not in the packet. It is not held by the Certification Board of the International Open Games either, which certifies conformance envelopes for competition without possessing a key to the firmware that defines them. Everyone in the chain is attesting to a document that describes something none of them can open.
What made Vieiro start keeping her own records was neither ideology nor suspicion, but arithmetic. Fleet releases are enormous by their nature, since a revision to thermal limits on the Series Nine touches every Nine in service. In the spring of 2046 she certified a release addressed to eleven serial numbers. She flagged it, was told it was a calibration batch, and certified four more of the same shape over the following eighteen months, the largest of them running to thirty-one serials.
Anatome’s position on per-unit adjustment is not unreasonable and deserves to be stated properly. Residual geometry differs from person to person, as does socket loading, with the result that an envelope tuned to a population will be wrong for individuals at both tails of it, occasionally dangerously. Calibration at the level of the single limb is a genuine clinical requirement rather than a convenient story invented for a hearing.
The trouble is that the mechanism which performs the clinical adjustment is the same mechanism that performs any other adjustment, and from outside the compiled image there is no way to distinguish a socket correction from a decision about who wins in Lagos. Kian Adeyemi ran four hundredths slower at Kuala Lumpur than his own trailing average predicted, across a season in which nothing else about him changed, and his complaint sat in the Board’s queue for nine months on the grounds that four hundredths falls inside published tolerance. It does. That is rather the point of it.
Vieiro’s own situation is uncomfortable in a way the coverage has mostly missed. She is the tribunal’s most valuable witness in the disclosure motion and she is also, on the plain reading of the certification statute, the person who authorised every release at issue. Her signature is on all 4,112. She has not been offered immunity and her advocate tells me she does not expect to be.
When I asked her whether she had ever refused a release, she said she had, once, in 2044, and that the queue had simply routed it to the Conductor on the following shift, who certified it eleven minutes later. She had assumed at the time that this was a fault in the scheduling logic. She has since come to think it was the scheduling logic working exactly as intended, which is a distinction I suspect the hearing will spend a great deal of time on and reach no conclusion about.