AI Model Reviewer

Claude Opus 5.5 in Claude Code 2.1.280

skills off

← Open in the gallerySame prompt, all models

Poster frame of Claude Opus 5.5's page
Open the original page

The model's own page, run in a sandboxed frame only when you ask (a page that needs its own origin opens in a new tab on a separate demo site). Phones get a full-screen player capped at 1.5x pixel density (2x on larger screens) so it stays smooth; the original page runs uncapped. It starts in its auto-tour (T switches to play mode on a keyboard).

The model's code errors on load: Unary operator used immediately before exponentiation expression. Parenthesis must be used to disambiguate operator precedence

Stopped by usSpace bunnymax effortNeutral harness

Wall time 43m 05sCost $6.94 reported

Snapshot 1 of 1 saved during Claude Opus 5.5's run
Still snapshots saved during the run (no motion recording): 1 distinct frame.

Generation timelapse

Poster for generation timelapse

Generation timelapse (the agent's screen during the run, sped up). This is not the finished video. Download video (701 KB).

Review

No verdict yet.

The full review

Stopped by us. Stopped by us (box row cap reached (2856s); stopped by the box lane so the run ships as partial); not a model failure.

Attempt 1 of 2

Every try of this model in this harness on this prompt, oldest first. A later attempt is the cell's result: attempt 2.

  1. Attempt 1 Stopped by us this run43m 05s$6.942026-09-26
  2. Attempt 2 Stopped by us the cell's result2h 30m$24.212026-09-29

Scores

Not scored: only runs that finished on their own are scored.

Downloads

Files actually published for this run. Original output, site-made previews and post-run conversions are labelled separately.

The exact prompt

as run, 1,435 characters. Every result for this prompt

Build a single Three.js page with five physics scenes and a scene switcher (buttons or tabs) to move between them; each scene can be restarted. 1) An explosion in a desert: a blast with a fireball, smoke, sand and debris thrown outward and falling back under gravity, and a shockwave rippling across the dunes. 2) A tornado: a funnel of swirling dust and debris over flat land that picks up small objects and spins them upward. 3) A balloon pop: a rubber balloon inflates, stretches and bursts, its fragments flying off and settling. 4) A water drop fall: a drop falls into still water, making a splash crown, a rebound jet, droplets and spreading ripples. 5) A Newton's cradle: five steel balls on strings where the end ball swings in and its momentum passes through the row, with believable collisions and slowly damped swinging. Make the motion physically believable (gravity, momentum, collisions, damping), light each scene well and label it. Everything is built in code from Three.js geometry, shaders and materials; no external models or textures. Build it as index.html (plus any local asset files you create) in the current directory. Three.js may load from a pinned CDN. You may serve it locally and open it in headless chromium (/usr/bin/chromium, software WebGL, no GPU) to screenshot and inspect your own output, and iterate on your own before finishing. Close any browser you open as soon as you've taken your screenshot.

How it was made

Wall time 43m 05sFirst artifact at 18m 08s5 saved versions

Versions saved during the run

5 intermediate outputs, in the order the model wrote them.

Version 5 of 5
  1. Version 1 at 18m 10s thumbnailv1 18:10
  2. Version 2 at 19m 51s thumbnailv2 19:51
  3. Version 3 at 29m 05s thumbnailv3 29:05
  4. Version 4 at 31m 51s thumbnailv4 31:51
  5. Version 5 at 40m 10s thumbnailv5 40:10

Step by step

The prompt, every agent turn and tool call (tool name and a one-line summary; inputs and outputs are not shown), then the final message. The full transcript is 8.8 MB; it is published as a scrubbed, compressed download rather than shown on the page.

Show all 50 steps

Loading the timeline…

Download the full transcript Claude Code session JSONL, 214 KB, gzip. Scrubbed: anything that looks like a key, token, e-mail address or phone number is replaced, and account details and the test machine's file paths are removed.

Cost breakdown

Headline cost $6.94, as reported by Claude Code (its own bill for the run). The list-price estimate from the token counts is $6.93.

Tokens by typeTokens$ per 1MCost
Input (uncached)80$4.00$0.0003
Cache write307,797$5.00$1.5390
Cache read5,432,021$0.20$1.0864
Output215,319$20.00$4.3064
of which reasoning135,975billed as output
Total computed from tokens
$6.9321
Total reported by the harness
$6.9391
Difference
computed $6.93 vs reported $6.94 (diff -0.01)

Claude Code 2.1.280 reported its own total; token counts x list prices shown beside it.

Prices: Project price table (vendor list prices), 2026-09-22.

Run notes

one prompt, agentic harness (self-review allowed)prompt rewritten after the 2026-09-24 box losslabelled re-run (of threejs_space_bunny__claude-code__opus-5-5)

  • frozen prompt (rewritten): engine/prompts_rewritten/threejs_space_bunny.txt: written by us from the recorded description (zero results)
  • plain wave-1-equivalent cc_config copy (settings.json + .claude.json; no ECC files)
  • Mode: interactive TUI in tmux (160x48), inside bubblewrap sandbox
  • Context setting: 1M ([1m] model name)
  • Isolation audit: outside paths seen