fix(reply): a member's link never loses room to the receipt

coilyco-ops commented

2026-08-13 12:28:46 +00:00

Member

403 gave the disclosure footer a send budget and made the issue reference the
casualty instead, because references are appended first and the budget shortens
whatever precedes the footer. Measured on main: at the ceiling the reference
goes and the footer stays.

That is a regression I introduced against behaviour that predates the footer.
References always survived before, and I traded an existing member-facing
guarantee for a feature shipped an hour earlier. Restoring the older one is not
a product judgement, which is how I framed it on 385 and on 413. It is the
conservative repair of a regression, and I should have made it rather than
asking.

The budget is withheld only when a reference was actually added, so the footer
still gets its full budget on every turn with nothing to protect, which is the
common case and the one 385 was about. Strictly better than both prior states
and no worse than either.

Deliberately not the convergent fix on 413. That one keeps both suffixes by
resolving references against the truncated answer, and it needs care I do not
have left today: truncating can remove a short reference the suffix resolves,
so a naive version ships a link to something the member cannot see. 413 keeps
the acceptance criteria and stays open.

closes #417

Co-authored-by: Kai Siren coilysiren@gmail.com
Co-authored-by: Claude noreply@anthropic.com
Co-authored-by: Kai Siren coilysiren@gmail.com
Co-authored-by: Claude noreply@anthropic.com

403 gave the disclosure footer a send budget and made the issue reference the casualty instead, because references are appended first and the budget shortens whatever precedes the footer. Measured on main: at the ceiling the reference goes and the footer stays. That is a regression I introduced against behaviour that predates the footer. References always survived before, and I traded an existing member-facing guarantee for a feature shipped an hour earlier. Restoring the older one is not a product judgement, which is how I framed it on 385 and on 413. It is the conservative repair of a regression, and I should have made it rather than asking. The budget is withheld only when a reference was actually added, so the footer still gets its full budget on every turn with nothing to protect, which is the common case and the one 385 was about. Strictly better than both prior states and no worse than either. Deliberately not the convergent fix on 413. That one keeps both suffixes by resolving references against the truncated answer, and it needs care I do not have left today: truncating can remove a short reference the suffix resolves, so a naive version ships a link to something the member cannot see. 413 keeps the acceptance criteria and stays open. closes #417 Co-authored-by: Kai Siren <coilysiren@gmail.com> Co-authored-by: Claude <noreply@anthropic.com> Co-authored-by: Kai Siren <coilysiren@gmail.com> Co-authored-by: Claude <noreply@anthropic.com>

coilyco-ops added 1 commit

2026-08-13 12:28:46 +00:00

fix(reply): a member's link never loses room to the receipt

ci / image-build (pull_request) Successful in 20s

Details

ci / test (pull_request) Successful in 6m14s

Details

ci / publish-echo-image (pull_request) Has been skipped

Details

ci / publish-observed (pull_request) Has been skipped

Details

c71f1ff7f4

403 gave the disclosure footer a send budget and made the issue reference the
casualty instead, because references are appended first and the budget shortens
whatever precedes the footer. Measured on main: at the ceiling the reference
goes and the footer stays.

That is a regression I introduced against behaviour that predates the footer.
References always survived before, and I traded an existing member-facing
guarantee for a feature shipped an hour earlier. Restoring the older one is not
a product judgement, which is how I framed it on 385 and on 413. It is the
conservative repair of a regression, and I should have made it rather than
asking.

The budget is withheld only when a reference was actually added, so the footer
still gets its full budget on every turn with nothing to protect, which is the
common case and the one 385 was about. Strictly better than both prior states
and no worse than either.

Deliberately not the convergent fix on 413. That one keeps both suffixes by
resolving references against the truncated answer, and it needs care I do not
have left today: truncating can remove a short reference the suffix resolves,
so a naive version ships a link to something the member cannot see. 413 keeps
the acceptance criteria and stays open.

closes #417

Co-authored-by: Kai Siren <coilysiren@gmail.com>
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Kai Siren <coilysiren@gmail.com>
Co-authored-by: Claude <noreply@anthropic.com>

coilyco-ops requested review from coilysiren

2026-08-13 12:28:46 +00:00

coilyco-ops merged commit cb360f080b into main

2026-08-13 12:56:37 +00:00

coilyco-ops referenced this pull request from a commit

2026-08-13 12:56:39 +00:00

Merge pull request 'fix(reply): a member's link never loses room to the receipt' (#418) from fix/a-reference-never-loses-to-the-footer into main

coilyco-ops referenced this pull request

2026-08-13 13:00:49 +00:00

Two service-authored suffixes are budgeted against each other, so one is always silently lost at the ceiling #413

Rows
Columns

fix(reply): a member's link never loses room to the receipt #418