# Pending question newline evidence

Both captures use the real web app in Chromium at 1280×800, light theme, fresh isolated server state, and the same synthetic replay ACP elicitation. The images are identical 700×440 crops at x=280, y=340 of that viewport. Two cropped screenshots 500 ms apart were pixel-identical before each final capture. No live provider account was used.

Base: e0d0fb33dd438dbe40737683ddd8224ea581ec08
Head: 9f3547e08ad78aaa71545092ed8bf6902c2c242b

The browser driver creates a project and a pending question through the replay provider, opens the seeded thread, and captures its expanded question. The question contains a Markdown heading, First line followed by one literal newline and Second line, and a bulleted list. Base displays both lines together. Head displays separate lines, preserving the heading and list.

## Re-run

Use the scenario harness from current main with dependencies installed. Put pending-question-newlines.e2e.ts and pending-question-seed.ts under scripts/scenarios. In scripts/scenarios/start-app.ts, import seedQuestion from ./pending-question-seed.ts after the shebang. Immediately after prepareHome(input.home), call:

```ts
yield* Effect.promise(() => seedQuestion(checkout));
```

This prepares the synthetic replay recording before the server's file watcher starts. It does not change application rendering code. Use separate base and head checkouts with dependencies installed. From the harness's scripts directory, run the same driver for each checkout, with T3_SCENARIO_APP_DIR pointing to that checkout:

```sh
T3CODE_DEV_SEED=0 E2E_TELEMETRY_DISABLED=1 T3_SCENARIO_APP_DIR=<checkout> node_modules/.bin/e2e run scenarios/pending-question-newlines.e2e.ts --target web --video off --output .e2e/<side>
```

ImageMagick convert is required for the identical crops and pixel comparison. Base fails the vertical line-position assertion; head passes. These assertions supplement the visually inspected captures.

AI-assisted; verified with the tests and tools listed above.

## Native iOS captures

The same seeded question was opened in the native dev client on pooled iPhone 18 Pro simulators, light theme, 402×874 screenshots. Images use identical 374×445 crops at x=14, y=398. Two cropped images 500 ms apart were pixel-identical on each side. Base joins the two lines; head separates them. Run pending-question-ios.e2e.ts with --target ios --video off --action-timeout 120000, using the same seed setup and per-side checkout selection. The head run reached its screenshot but its original whole-file hash check failed; subsequent ImageMagick pixel comparison showed only one differing pixel outside this crop and zero inside. The base driver uses the corrected pixel comparison.