ME.Preface:7.6 - Practical checks, recurring failures and costs

Recognition and assurance answer different questions. First identify the missing result and a plausible contribution. Then check only the claims on which the actual use relies, with the evidence and independence that their conclusion needs. Do not treat finding the right heading, obtaining a graph path or reading all patterns as assurance.

For a production result, ask whether the receiving action and stop are recoverable; whether sources and candidate evidence retain their roles; whether each merge, split and omission has a reason; whether the language keeps Method identity separate from body allocation; and whether a changed premise has an identifiable return. When the promised use spans several patterns, challenge at least the material relation on which that use depends. Broaden a check when the dependency boundary is unknown, not merely because the language is large.

The recurring failures have distinct repairs. A chapter copied as a Method returns to recovery and qualification. A tidy graph that loses a condition returns to description content. A comparison that changes both content and form returns a bundled result, not a form advantage. A coverage table that counts topics returns to reconstruction. A source-access failure stops a required new recheck; it does not retrospectively falsify everything derived from the source.

Source familiarity, preference for a neat architecture and enthusiasm for pattern languages can bias all these judgments. Preserve the rival account and the cheaper direct route. AI-assisted recovery or comparison retains its model, input and interaction limits when they matter to the claimed result; fluency does not replace source access or independent use evidence.

The practical benefit is reusable source decisions and more local reconsideration. The burden is additional initial reading and maintenance of the few dependencies needed for return. In one bounded comparison with two clean-context AI readers, both the full situated-use body and a truthful entry-only alternative supported the required continuation. After a changed condition, the full-body reader reused its return guidance without another source opening; the entry-only reader reconstructed that boundary from two further source passages. This is a local reconstruction benefit, not a measured total-effort, correctness or human-performance advantage. The full body also requires its upfront reading and maintenance. Whole-production effectiveness remains unestablished. Use a simpler account when maintaining the language does not repay its burden.