Japanese garden

One picture, three 3D worlds. A new picture, Claude Opus 5.5 at three effort levels.

Claude Opus 5.5 in Claude Code built a new AI-generated picture, a miniature Japanese garden, as a Three.js world on medium, high and xhigh effort. The prompt is Lakeside build's with the garden's scene nouns, the same sun and wind sliders and the same copy of Three.js. One run each, with no human edits.

This is not a CmdBench benchmark run: no checks, scores or verdicts.

AI-generated picture of a miniature Japanese garden on a round stone base: a wooden temple with glowing lanterns, a stone lantern, raked gravel with rocks, pine trees and a red maple, a bamboo fence, and a red arched bridge over a pond with koi, in warm light.
Reference picture, AI-generated. Every run got this picture and the same prompt.

The three worlds

One run at each effort level. Each world is served exactly as the agent wrote it, with no fixes, and opens full screen in this tab. The pine crowns in the high build render solid black in every view: that is the build as written.

  • Frame from the medium-effort build, captured from the build itself. Opens the world.

    Effort: medium

    Built by
    Claude Opus 5.5, medium effort
    Time
    25 min 51 s
    Output tokens
    162,486
    Cost estimate
    $5.12

    Drag to orbit, scroll or pinch to zoom, and move the Sun and Wind sliders.

  • Frame from the high-effort build, captured from the build itself. Opens the world.

    Effort: high

    Built by
    Claude Opus 5.5, high effort
    Time
    31 min 26 s
    Output tokens
    200,533
    Cost estimate
    $6.31

    Drag to orbit, scroll or pinch to zoom, and move the Sun and Wind sliders.

  • Frame from the xhigh-effort build, captured from the build itself. Opens the world.

    Effort: xhigh

    Built by
    Claude Opus 5.5, xhigh effort
    Time
    48 min 48 s
    Output tokens
    306,813
    Cost estimate
    $12.18

    Drag to orbit, scroll or pinch to zoom, and move the Sun and Wind sliders.

Costs are Claude Code's own estimates at API list prices.

How it was made

Picture
AI-generated. The exact PNG every run got
Prompt
The same for every run: read the prompt (TASK.md)
Model
Claude Opus 5.5 (claude-opus-5-5)
Agent
Claude Code 2.1.319
Effort levels
medium, high, xhigh, one run each
Tools
File tools only: no browser, no shell, no web. The agent never saw its world running.
Time limit
60 minutes per run
Three.js
0.187.0
Human edits
None. Each world is served exactly as the agent wrote it, with no fixes.
Run date