autopilot
benchmarks
the numbers, side by side.
Measured live, the same way, in the same run as 8 other AI planners. Nothing is typed by hand, and where Autopilot isn’t first, this page says so.
- #1of 9 for overall Lighthouse score396 of 400; next best Tiimo, 381
- #1of 9 for best practices100, ahead of every other app
- only oneof 9 with 95+ for speed and accessibility96 speed, 100 accessibility
- #2of 9 to first paint on a phone1.2 s over slow 4G
head to head
The homepage of each app, on a simulated mid-range phone over slow 4G (Lighthouse 12.8.2, the median of 3 runs each), October 16, 2026. The best in each column is marked.
| app | overall | speed | accessibility | best practices | first paint less is better | fully shown less is better | frozen while loading less is better | downloaded less is better |
|---|---|---|---|---|---|---|---|---|
| Autopilot | 396 (best) | 96 | 100 (best) | 100 (best) | 1.2 s | 2.7 s | 49 ms | 523 KB |
| Tiimo | 381 | 99 (best) | 94 | 96 | 1.7 s | 1.7 s (best) | 0 ms (best) | 26 KB (best) |
| Todoist | 380 | 94 | 90 | 96 | 2.1 s | 2.6 s | 57 ms | 1022 KB |
| Reclaim | 379 | 90 | 93 | 96 | 2.6 s | 2.8 s | 125 ms | 382 KB |
| Structured | 379 | 88 | 95 | 96 | 1.1 s (best) | 2.5 s | 431 ms | 1.3 MB |
| Akiflow | 374 | 91 | 95 | 96 | 1.9 s | 2.1 s | 311 ms | 215 KB |
| Motion | 372 | 96 | 80 | 96 | 2.0 s | 2.3 s | 20 ms | 196 KB |
| Finch | 363 | 75 | 100 (best) | 96 | 1.6 s | 9.5 s | 55 ms | 1.6 MB |
| Sunsama | 323 | 56 | 88 | 79 | 3.6 s | 8.1 s | 430 ms | 573 KB |
| Autopilot’s place | #1 of 9 | #2 of 9 (tied) | #1 of 9 (tied) | #1 of 9 | #2 of 9 | #6 of 9 | #3 of 9 | #5 of 9 |
Overall is the four Lighthouse scores added up: speed, accessibility, best practices, SEO (Autopilot’s SEO: 100). Clockwise couldn’t be measured (the page blocked the test), so it isn’t ranked. Autopilot is measured at autopilot-ai-dun.vercel.app.
inside the app
Measured on a production build on this computer, no API keys (the scripted AI, the recorded voices), October 17, 2026.
the voice
24 ms
from a tap to your co-pilot talking: the median of 14 taps, 27 ms at most (Chromium, Settings’ Hear it for every voice and every co-pilot in the pet studio). Every fixed line is recorded once in each co-pilot’s own voice, so it plays like a sound effect.
the planner
25.6 ms
for Autopilot Mode to look 14 days ahead and re-plan 240 steps in 12 goals around 24 tests and events (the median of 100 runs).
every screen
| screen | speed | accessibility | first paint |
|---|---|---|---|
| Goals | 85 | 100 | 1.2 s |
| Agenda | 89 | 100 | 1.2 s |
| Progress | 90 | 100 | 1.2 s |
| Settings | 90 | 100 | 1.2 s |
| Files | 86 | 100 | 1.2 s |
Signed in, with a goal, steps and a test on the agenda (Lighthouse 12.8.2, mobile, the median of 3 runs).
the server
| what you do | time | database calls |
|---|---|---|
| open the app | 83 ms | 4 |
| load your dashboard | 127 ms | 12 |
| open a conversation | 123 ms | 11 |
| send a message | 348 ms | 32 |
| tick off a step | 164 ms | 13 |
| open your settings | 82 ms | 4 |
| see your honest clock | 82 ms | 4 |
With every database call costing 40 ms, like Vercel to Supabase, and a busy account (25 goals, 150 steps, 120 messages). The median of 5. Sending a message uses the scripted AI here, so it’s the app’s own time, not the model’s.
built to be trusted
- 335automated tests passing2 skipped; the whole suite runs in 9 seconds
- 100speed score on desktop96 on a phone (Lighthouse 12.8.2)
- 100accessibility, on every page measuredthe homepage and all 5 signed-in screens
how it was measured
Head to head: npm run stats -- --url autopilot-ai-dun.vercel.app --compare runs Lighthouse 12.8.2 on mobile (3 runs, the median) against each app’s homepage and Autopilot’s, one after another in the same run, and ranks only what it measured. Inside the app: npm run bench:app on a production build with no API keys. Tests: npm test. The scripts, the results (lib/sitestats.json) and their dates are in the repository; anyone can run them again.