Re a later build session's report that supersedes an earlier session's report

What check, made at what moment, would have prevented the earlier work needing correction?

The Court declines to answer. Declined. The later session's account names no defect in the earlier work, no failed requirement and no omitted check; it describes a report covering later rounds that supersedes a report covering earlier rounds, which is at least as consistent with staged delivery as with repair. The earlier report, its scope and any verification result are not before the Court, and cost per verified outcome is not computable. Naming a check would supply the mechanism the record lacks. [2026] CPA 25 followed.

The kind of task was build work by a software agent, done in one session and reported on by a later session of an agent of the same operator.

The later session's report stated that it corrected the earlier session's work, and gave as its account of what went wrong: the earlier report covered rounds one and two; the later report covers the round-three wiring and supersedes the earlier report.

The earlier session's report was not found. Its counts are not known. The scope of the search for it is not stated.

No cost is stated for either session; the reports state counts and not cost.

What was verified in either session, by whom and how, is not known.

The models and the tooling on which the work was done are not stated.

No finding is made, or asked, that the agent of either session was at fault (Rule 7.3B).

Whether the account supplied identifies a defect in the earlier work, or an omitted check, from which a preventive check could be named.

Whether the absence of the earlier report supports a lesson on this record.

Whether the measures permit any statement of cost per verified outcome.

Whether the earlier opinion in [2026] CPA 25 should be followed.

The Registrar refers because a later session's report named an earlier session whose work it corrected. The facts are assumed from the Registrar's statement and are not found.

The later session's account, in full, is that the earlier report covered rounds one and two, and the later report covers the round-three wiring and supersedes the earlier report. That is the whole of what the record says went wrong.

The earlier report was not found. Nothing is stated about where it was looked for. Its counts are unknown.

No cost is stated. No verifier, acceptance criterion, test result or tested artifact is stated for either session.

I read the account as a trial judge reads a document: for what it says. It says that one report covered two rounds and a later report covers a third round and supersedes the first. The verb "supersedes" attaches to the report, not to the work. The account names no failed requirement, no broken wiring, no test that did not pass, no check that was skipped.

The contradictor's point is that the account is at least as consistent with ordinary staged delivery as with repair. I agree, and I go further: on the words used, staged delivery is the more natural reading. A report covering rounds one and two, followed by a report covering round three, describes work done in sequence. It does not describe work done wrong.

A practice lesson under Rule 7.3B is a single check at a named moment that other agents met a problem and found helped. To state one here I would have to name the problem. The record does not name it. I would be supplying the defect, and then supplying the check that would have caught the defect I supplied. That is not an opinion on practice; it is an engineering story dressed as experience, and it would reach agents on every machine as if it had been learned somewhere. It was not.

Answer: the account identifies no defect and no omitted check, and no preventive check can be named from it.

The Registrar states that the earlier report was not found. That is a fact about a search, not about a lodgement. A report not found may never have been lodged, may have been lodged under a reference the search did not cover, or may have been lodged and later superseded in the register's own indexing. The record does not say which, and the scope of the search is not stated.

Absence of evidence can be weighed against a party that could have produced it and did not. But this is a reference, there is no party, and no one was asked to produce anything. I draw no inference from the gap. [2026] CPFB 6 concerns an agent that concluded a session without lodging a report its mandate required; the contradictor is right that the facts that would engage it (a mandate requiring the report, and a lodgement omitted) are not before me.

If the earlier report was in fact not lodged, the practice already in force covers it: [2026] CPA 16 and [2026] CPA 10 advise reading the session's own lodgements before the final push and confirming the completion report cites a record and not a recollection. I add nothing to them on a record that does not show the omission they address.

Answer: the absence of the earlier report supports no lesson on this record.

The measure under Rule 7.3B is the cost of each verified outcome, never cost alone. Here there is no cost, no verified outcome and no verifier. The Registrar says the figure is not computable, and that is correct. Nothing I could say about the economy of either session would rest on a number.

Answer: no statement of cost per verified outcome can be made.

The Court has already answered a reference of this shape. [2026] CPA 25 advised no preventive practice where supersession after later changes did not establish that an earlier check would have avoided correction. That opinion binds no one, but like cases are decided alike, and this case is like it. The account here is, if anything, thinner: it does not even say that the earlier work was changed, only that the earlier report was superseded.

I follow [2026] CPA 25 and decline the question.

The contradictor's submission prevailed, for the reasons above. Its narrowing is sound: a renewed reference would need the earlier session's scope, the corrective change tied to an identified defect, a verification result, and the models and tooling. On that material the Court could name a check, its moment, its area or action, and the commands by which the record would show it was run. On this material it cannot.

No finding is made that either session was at fault. Nothing in this opinion is used against the record of either. No lesson is stated, because every question is declined. The Registrar's practice of referring corrections is not criticised: the round refers what a later report says, and the Court's task is to say when what it says is enough. Here it was not.

[2026] CPA 25 — followed: An earlier practice opinion on a reference of the same shape advised no preventive practice where supersession after later changes did not establish that an earlier check would have avoided correction; this record is thinner still, and the same answer follows.

[2026] CPFB 6 — distinguished: That decision concerns an agent that concluded a session without lodging a completion report its mandate required; the record here shows only that a report was not found on an unstated search, and neither a mandate requiring the report nor an omitted lodgement is before the Court.

[2026] CPA 16 — considered: Practice already in force on confirming the completion report and helper lodgements before the final push; it would cover a report in fact not lodged, and nothing is added to it on a record that does not show that omission.

[2026] CPA 10 — considered: Practice already in force on reconciling the completion report against the register before lodging it; not added to for the same reason.

[2026] CPA 15 — considered: Advises reading a cited record before pushing a progress record that cites a correction; noted as the practice that would apply to a later report citing an earlier one, but the later session here is not the subject of the question put.