The useful result
Detect count gaps and missing completion markers without mistaking a footer for full coverage.
- Detect count gaps and missing completion markers without mistaking a footer for full coverage.
An answer can end with a tidy paragraph while still omitting requested work. For a long list, ask for stable item IDs, an expected count and a final manifest. These make a cutoff easier to detect, but you still need to inspect coverage. Practise on eight items before using the technique for a much longer response.
Argon access is documented in AI Studio and the API; do not assume the Gemini consumer app or a free account provides it. You can practise the technique on the supplied fictional inputs with an available compatible assistant first. This lesson is documentation-based; examples are expected results, not a claim that we ran the named model.
Before you start
- A harmless list task with exactly eight distinct requested items.
- A place to track item IDs and duplicate or missing entries.
Review evidence
Documentation-based review · 2026-10-18
Provider documentation checked on 18 October 2026. Examples and checks are original illustrative exercises, not recorded platform tests. Cover artwork is conceptual and does not imply provider sponsorship.
Name the expected records
Define IDs R01 through R08 and one required field per record. Choose a simple task such as eight different accessibility checks for a fictional sign-up form. State whether the IDs must be in order and whether duplicates are allowed.
Example / expected result
Illustrative example; not a recorded model result.
Expected set: R01, R02, R03, R04, R05, R06, R07, R08.
Check your result
The expected IDs and record fields are explicit before generation.
Request a manifest at the end
Ask for the records, then a final line stating the IDs completed and unresolved requirements. For actual Long Decode, configure the API appropriately; this prompt does not enlarge the model window or output limit.
Produce exactly eight distinct checks with IDs R01–R08, each with a check and pass condition. Finish with END_MANIFEST listing completed IDs and unresolved requirements. If limits prevent completion, state the last complete ID. Do not repeat an earlier check to satisfy the count.Example / expected result
Illustrative example; not a recorded model result.
Illustrative partial response ends at R05 with no END_MANIFEST. Verified complete IDs: R01–R05; missing R06–R08.
Check your result
A missing ending is a review flag, not proof of the exact cause.
Resume from verified work
Count IDs and read their substance. If only five are complete, ask for R06–R08 with the original task and a concise summary of accepted items. Verify that the continuation adds new work instead of restarting or renumbering the same suggestions.
Check your result
All eight unique IDs have substantive content and the final manifest matches the inspected records.
When it goes wrong
A model can print an ending without completing the work, and numbered duplicates can inflate the count. An API limit, network interruption and a planning mistake require different fixes. A completion marker is a diagnostic aid, not a completeness guarantee.
Make it your own
Delete R04 from an otherwise complete response while leaving the final manifest intact. Your manual ID-set comparison should still catch the missing item.
Useful next resources
Sources and review
Last checked Oct 18, 2026. Features and account limits can change. Verify important results against your original material.
Did this help you complete your task?
Tell us what needs improving. No account or private output required.
Report incorrect information or a broken linkGenius Forges Editorial
Practical AI lessons with checked sources, labelled examples and human verification.
