New version 0.5.7

You have 0.5.5. Found when this page was made.

What's new:

Every draft error is listed at once; a stdin draft is kept for line edits
A table with no |---| row is an error instead of plain text
Long flow edges are no longer cut off at the edge of the diagram

Run in a terminal:

npx skills update answer-me-with-html -y

Or tell your agent:

Update answer-me-with-html

m.allternit.com slowness check

What I measured (2026-10-22)

The server is fine. The slowness is on the phone: too much code at startup and too many animations running.

Check	Result
Server response	~250 ms, assets compressed (br/gzip)
First load	416 requests, 5.6 MB of JS, ~6 s on a mid phone (4× CPU slowdown, 9 Mbps)
Startup work	one 830 ms freeze during boot
Sign-in screen, sitting idle	69 CSS animations running; CPU busy 43% of the time
First-run setup screen, idle	33 animations; a layout pass every frame (60/s)
Phone home with nothing loaded	quiet (1 animation)

Causes, ranked

The phone entry loads desktop-size code before it shows anything. The HTML preloads ~340 modules. They include Project home, the design registry (496 KB), charts and three.js. The count was 328–341 for days. So this is old weight, not a new break.
The pixel orb animates every square on its own. 22–33 SVG rectangles each run a CSS animation. Phones can't hand these to the GPU, so they redo styles every frame. The sign-in showcase adds about 35 more.
Not reproduced yet: the signed-in screens with real data. A dev login reaches only setup screens, so bot stages, chats and polling with a real account are still unmeasured.

What I need from you

[!ask] Which screen is slow (home, a bot chat, Bots, Code, ACI), on which phone, and did it start today? Or set ALLTERNIT_TEST_EMAIL / PASSWORD / OTP so I can profile a real signed-in session.

Proposed fix (needs your go)

Lazy-load the heavy views in the phone entry so first paint loads only the shell.
Animate the orb as one element, or pause it when not visible; follow reduced-motion and battery saver.
Deploy to m.allternit.com only after your go (it's a manual wrangler step).

---
title: m.allternit.com slowness check
---

# What I measured (2026-10-22)

The server is fine. The slowness is on the phone: too much code at startup and too many animations running.

| Check | Result |
|---|---|
| Server response | ~250 ms, assets compressed (br/gzip) |
| First load | 416 requests, 5.6 MB of JS, ~6 s on a mid phone (4× CPU slowdown, 9 Mbps) |
| Startup work | one 830 ms freeze during boot |
| Sign-in screen, sitting idle | 69 CSS animations running; CPU busy 43% of the time |
| First-run setup screen, idle | 33 animations; a layout pass every frame (60/s) |
| Phone home with nothing loaded | quiet (1 animation) |

# Causes, ranked

1. **The phone entry loads desktop-size code before it shows anything.** The HTML preloads ~340 modules. They include Project home, the design registry (496 KB), charts and three.js. The count was 328–341 for days. So this is old weight, not a new break.
2. **The pixel orb animates every square on its own.** 22–33 SVG rectangles each run a CSS animation. Phones can't hand these to the GPU, so they redo styles every frame. The sign-in showcase adds about 35 more.
3. **Not reproduced yet: the signed-in screens with real data.** A dev login reaches only setup screens, so bot stages, chats and polling with a real account are still unmeasured.

# What I need from you

> [!ask]
> Which screen is slow (home, a bot chat, Bots, Code, ACI), on which phone, and did it start today? Or set ALLTERNIT_TEST_EMAIL / PASSWORD / OTP so I can profile a real signed-in session.

# Proposed fix (needs your go)

- Lazy-load the heavy views in the phone entry so first paint loads only the shell.
- Animate the orb as one element, or pause it when not visible; follow reduced-motion and battery saver.
- Deploy to m.allternit.com only after your go (it's a manual wrangler step).