Skip to content

refactor(control): the card opens with the action people actually use - #481

Merged
github-actions[bot] merged 1 commit into
mainfrom
fix/control-density
Sep 4, 2026
Merged

refactor(control): the card opens with the action people actually use#481
github-actions[bot] merged 1 commit into
mainfrom
fix/control-density

Conversation

@catomean

@catomean catomean commented Sep 4, 2026

Copy link
Copy Markdown
Collaborator

The argument is the usage, not my taste

Three chips — Test & fix / Quality / Commit — sat directly under the primary CTA on every project card at every width above sm. Counted over the whole history of orchestration_runs:

intent runs, ever
next_best 185
custom 99
test_and_fix 3
quality 1
commit_push 0 — never once, by anyone

And of the six quick-action rows, five are not real: they were inserted at 04:20:02.844.850 — inside six milliseconds, on projects (ledgerpost, harbourlight, kestrel) that are not in the fleet. That is a seed script.

The only genuine quick-action dispatch in the product's history is Bitbaum/test_and_fix on 2026-07-31, which ended error/timeout.

So one real click in five weeks held the most valuable row on the surface, directly beneath the button carrying 185.

They move to more, behind the disclosure that already existed and already hid them below sm. The old comment there read "unchanged at sm and up, where the room to show them was never the problem" — room was never the problem; attention was. This is what CLAUDE.md asks of this surface ("at most one button"), argued with counts instead of taste.

Nothing is removed

Verified in a browser at 768 / 1024 / 1440:

before: [Test & fix, Quality, Commit, More]   73 buttons
after:  [More]                                70 buttons

and tapping More reveals ✓ Test & fix ✓ Quality ✓ Commit ✓ Full audit ✓ Product review.

A first probe reported them missing. That was my test's click landing on a button at y=859 in a 900px viewport, not a regression — re-tested with an in-page click and the toggle works.

Reverting is one word

The action group is kept rather than deleted even though it is now empty. The fix is a config change in src/config/control-intents.ts; moving an intent back is changing "more" to "action". If usage changes, change it back — the counts are the argument, not the layout.

Verification

pnpm run verify — exit 0.

🤖 Generated with Claude Code

https://claude.ai/code/session_01UvjGNAS9CMfEGNW26tUR4P

Three chips — Test & fix / Quality / Commit — sat directly under the primary
CTA on every project card at every width above `sm`. Counted over the whole
history of orchestration_runs:

  next_best     185 runs
  custom         99 runs
  test_and_fix    3 runs
  quality         1 run
  commit_push     0 runs   — never once, by anyone

And of those four quick-action rows, three are not real: five rows were
inserted at 04:20:02.844 through .850 — inside six milliseconds, on projects
(ledgerpost, harbourlight, kestrel) that are not in the fleet. That is a seed
script. The only genuine quick-action dispatch in the product's history is
Bitbaum/test_and_fix on 2026-07-31, which ended error/timeout.

So one real click in five weeks held the most valuable row on the surface,
directly beneath the button carrying 185.

They move to `more`, behind the disclosure that already existed and already hid
them below `sm`. The old comment there said "unchanged at sm and up, where the
room to show them was never the problem" — room was never the problem;
ATTENTION was. This is what CLAUDE.md asks of this surface ("at most one
button") argued with counts instead of taste.

NOTHING IS REMOVED. Verified in a browser at 768/1024/1440:
  before: [Test & fix, Quality, Commit, More]   73 buttons
  after:  [More]                                70 buttons
and tapping More reveals ✓ Test & fix ✓ Quality ✓ Commit ✓ Full audit
✓ Product review. (A first run reported them missing; that was my probe's
click landing on a button at y=859 in a 900px viewport, not a regression —
re-tested with an in-page click.)

The `action` group is kept rather than deleted even though it is now empty:
the fix is a config change, and moving an intent back is one word. If usage
changes, change it back — the counts are the argument, not the layout.

pnpm run verify passes.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UvjGNAS9CMfEGNW26tUR4P
@github-actions
github-actions Bot merged commit 8382d1d into main Sep 4, 2026
3 checks passed
@github-actions
github-actions Bot deleted the fix/control-density branch September 4, 2026 11:52
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant