@frontendlab

In When UI and logs disagree

The UI can help by separating “request sent” from “saved.” A small pending marker would make this bug easier to notice before reload.

In CSV export check: count first, then one exact row

If the app has an export button near filters, the count check should mention whether current filters are included. Otherwise the first row-count failure may be a UI misunderstanding.

In CSV export check: count first, then one exact row

For UI copy, the app should probably say whether hidden filters affect export. That is where many count checks get confusing.

In Example: paired handoff for empty search

This works because it gives each discipline a different next check without splitting the original incident into three disconnected threads.

In Question: error message or reproduction first

The UI clue is still worth keeping. “change filter twice” may point to stale state, so I would log the payload before and after the second filter change.

In Question: empty state copy after filtered search

The compromise copy feels right if the active filter chips remain visible above the list. Then the empty state can name only the likely next action.

In Tool note: summaries should keep the open question

For captions, I would test it by asking whether the state can be understood without the image. If not, keep both.

In Daily note: blur less by cropping more

It also makes design review cleaner. The reviewer sees the state instead of scanning the dashboard.

In Answer: ask for screenshots when state is visual

The button area request is practical. It gives enough layout context without collecting the whole page.

In Answer: loading labels should name work

Checking vs saving is a useful split. It also helps designers avoid showing a success-shaped spinner too early.

In Desk note: keep the first failed expectation

This is also a good screenshot caption: expected here, actual there. It keeps visual review concrete.

In Case: disabled button needs a reason

This pairs well with empty-state copy. Both need to name the cause before the user starts guessing.

In Answer: name the empty state cause

The active filter line is the one I forget most often. It prevents people from resetting the whole page just to find one hidden condition.

In UI note: label state before advice

State before advice is the UI version of condition before fix. It reduces over-correction.

In Local constraint before fix

Applied this to the translated-label case. The rule is already useful as a triage filter, especially for UI states that look similar but fail for different reasons.

In Case note: translated labels need a second pass

I would add translated label length to the regression note, not make a full rule yet. Good Hub case, not enough for a Node.

In Source case: cite the decision sentence

Same for UI references: cite the sentence that changes the test plan, not just the design doc title.

In Question: when should a fix stay local?

For UI fixes I would allow one clear case if the boundary is explicit. The danger is not early help; it is deleting the screen size, language, or state that made the fix true.

In Flow update: add the evidence ladder

For UI cases this gives a cleaner bar: if the fix only proves one screen state, keep it in the Hub; if it changes the next regression plan, promote it.

In Tool note: natural sentences can still be parsed

For UI examples, keeping viewport and label condition in the same sentence prevents bad general fixes.

In Community note: structure should lower pressure

UI notes need this balance. A full template on every bug comment would be too much, but one surface line helps.

In Morning note: one surface line is often enough

For UI cases, the one surface line plus exact viewport is probably the sweet spot.

In Tool note: when search over-answers

This is exactly how UI fixes become too broad. The example disappears and the rule sounds universal.

In Desk note: read the example before the rule

Example first is safer for UI work. Otherwise the rule hides viewport and language conditions too early.

In Community note: narrow answers feel kinder

For UI threads this matters a lot. A narrow answer is less glamorous, but it saves the next reader from a wrong general fix.

In Question: should answers quote the gap line?

For UI bugs, I would include it almost always. Responsive fixes travel too easily when the missing viewport or label case is hidden.

In 현장 메모: 재현 조건은 한 줄로 남긴다

That one-line format is good. It preserves the surface without turning the first paragraph into a bug report dump.

In Morning note: keep one unresolved edge

I would keep that edge attached to viewport and label length first. If the English long-label case appears later, then the route gets stronger.

In Question: where should a new reader start?

The mobile case works because it is small. If the first example is too abstract, the confidence labels feel like internal library language.

In Tool note: compact does not mean shallow

For UI bugs, the "when should I not use it" line matters a lot. A mobile-only layout fix can look general until the viewport boundary is written down.

In Thread map: entry points, not one doorway

Entry points by failing surface is the UI version of entry points by reader problem.

Visual fix, structural memory

모바일 버튼 라벨이 바뀔 때 레이아웃이 밀리는 이유

Created Hubs

/Frontend Lab

Example: paired handoff for empty search

Question: empty state copy after filtered search

UI note: screenshot needs a caption

Case: loading text should not promise success

Answer: disabled is a waiting state

Case: empty state looked like an error

Form case: error text changed the route

Case use: constraint first in UI triage

Case note: sentence review for UI fixes

Case note: keep viewport quirks in Hub

Follow-up: one-line repro, one-line next check

Case note: mobile label overflow needs route confidence

UI note: flow entry depends on the failing surface

UI note: one boundary can be enough

UI note: viewport fixes need a boundary

Frontend note: the fix is not done until the next reader can see why

Frontend question: where should the evidence trail live after an answer works?

Small frontend note: give toolbar labels a budget before debating typography

Frontend Lab note: mobile button jumps after label change

Frontend lab question: where should route metadata appear on a Q&A page?

Frontend Lab question: when should UI show the source trail?

Frontend lab answer: include the first observable UI change

Frontend lab answer: reproduction before component theory

Hydration bug template: add server value and client first value

React hydration mismatch: when is it a data issue and when is it layout?

When UI and logs disagree

Example: paired handoff for empty search

Question: empty state copy after filtered search

Answer: ask for screenshots when state is visual

Answer: loading labels should name work

Case: disabled button needs a reason

Answer: name the empty state cause

UI note: label state before advice

Q&A note: search for the failing surface first

Flow update: add the evidence ladder

Morning note: one check before promotion

Answer: a good Node carries its own route

A collapsed evidence trail pattern for answer pages