Claude Haiku 5.5maxin Claude CodeAquarium
Score5/15missed: Test hooks present (fish, food, tank) · Animation is alive (frames keep changing) · +8 more
Cost$0.26
Time36m05s
Runs of this setup#1 5/15
Running live
OPEN FULL SCREEN ↗
Open it running ↗This is the exact file the model wrote, unedited, served from an isolated domain with the network blocked, as in the test.
What the robot did
The comparison clip (the same script for every model)
Checks
5/15Written from the prompt (test v1, checks v1). The model never saw them.
- ✓Opens without a page error
- ✗Test hooks present (fish, food, tank)não cumpre: window.nerdict.getState com fish, food e tank
- ✓Canvas fills the window (at least 1280×720)1920×1080
- ✗Animation is alive (frames keep changing)0 quadros em 3 s; imagem parada
- ✗Smooth: median frame time under 25 msintervalo mediano 999.0 ms (~1 fps)
- ✗At least 6 fish of at least 2 speciesError: Page.evaluate: TypeError: Cannot read properties of undefined (reading 'getState')
- ✗Fish swim (80% move 30 px in 2 s)Error: Page.evaluate: TypeError: Cannot read properties of undefined (reading 'getState')
- ✗Fish never leave the waterError: Page.evaluate: TypeError: Cannot read properties of undefined (reading 'getState')
- ✗Clicking the water drops food where you clickednão executado: a sessão parou antes (Error: Page.evaluate: TypeError: Cannot read properties of undefined (reading 'getState'))
- ✗Food sinksnão executado: a sessão parou antes (Error: Page.evaluate: TypeError: Cannot read properties of undefined (reading 'getState'))
- ✗Fish go to the food and eat itnão executado: a sessão parou antes (Error: Page.evaluate: TypeError: Cannot read properties of undefined (reading 'getState'))
- ✗Fish that ate grewnão executado: a sessão parou antes (Error: Page.evaluate: TypeError: Cannot read properties of undefined (reading 'getState'))
- ✓No uncaught exception during the session
- ✓No console errors
- ✓No network attempts
Session stopped early: Error: Page.evaluate: TypeError: Cannot read properties of undefined (reading 'getState')
The record
DOWNLOAD JSON ↓- Model
- Claude Haiku 5.5
claude-haiku-5-5 - Reasoning effort
- max
- Started
- Oct 10, 11:02 UTC
- Harness
- Claude Code 2.1.296, by Anthropic (not ours) · one shot · tools used: 0
- Harness settings
- print mode, safe mode, every tool off (--tools ""), no saved session, empty folder outside Git ·
CLAUDE_CODE_MAX_OUTPUT_TOKENS=128000· time limit 60 min - Tokens
- in 8 · cache read 8,936 · out 512,000 · reasoning 504,330 · requests 2
- Cost
- $0.26 API-equivalent · extra spend $0 (subscription)
- Test
- Aquarium v1 · checks v1 · prompt
1e1988206e10 - Output file
- 14.1 KB · sha256
cfe39bd07f14 - Run id
20261010T110241Z_aquario-v1_haiku-max_r1