Give one prompt and 6 hours for a 3D visualisation
Same prompt to GPT-6 Astra in Codex and Opus 5.5 in Claude Code; Opus ran 6 subagents and finished a three.js piece in 1h25.
Prompt for your agent
Goal: Compare how models handle a long one-shot design task. Setup: agents used in source: Claude, Codex; tools/MCPs stated: three.js, pnpm. Steps: 1. Prompt: “Make a three.js (pnpm) visualization of all Invisible Cities by Italo Calvino. Don’t ask questions, it is a one-shot task. You have 6h of work, use it until it becomes a masterpiece.” 2. Run it in Codex (GPT-6 Astra) and in Claude Code (Opus 5.5). 3. Compare: Opus used 6 parallel subagents (~7 agent-hours) and stopped after 1h25. 4. Try the same prompt on other models and harnesses. Rules: ask me before anything irreversible (sending, posting, paying, deleting); don’t invent data, results or tools I haven’t given you; tell me what you couldn’t verify. Based on: Piotr Migdał — https://quesma.com/blog/invisible-cities-one-shot/
Steps are from the source. Includes a “Based on” credit line.
I tried this
Tell the next builder. One anonymous answer each; click again to undo.
Agents, models and tools only what the source states
Sources & credit
- Piotr Migdał quesma.com primaryblog · 7 Oct 2026https://quesma.com/blog/invisible-cities-one-shot/
- staredHacker News · 8 Oct 2026https://news.ycombinator.com/item?id=50004790