Long prompt or one line: inspect the racer and Lighthouse builds

Keep the two variables visible

The 30 April 2026 recording compares Claude Opus 4.7 and GPT-5.5 in Codex with verbose and single-line prompts. The two builds are a Synthwave 3D racing game and a landing page for the fictional focus app Lighthouse. Prompt length and model choice both change in the comparison, so record each combination before interpreting a result.

1. Start at 0:57 for the detailed racer brief and 3:27 for the one-shot racer comparison.
2. Keep the task, prompt version, model and starting state together in your notes.
3. Write separate criteria for functioning game mechanics and attractive presentation before reviewing either result.

Test the game and inspect the landing page

The description reports mixed outcomes: a short prompt sometimes wins, and the model Luke expected less from can produce a better-looking result. It also notes that a real photo made a substantial contribution to one Codex output. Those are observations about these examples, not proof that shorter prompts or one model are always better.

1. Play the racing result and check the requested mechanics instead of judging only its first frame.
2. Follow the detailed Lighthouse build at 6:47 and the short-prompt page at 9:00.
3. Compare layout, imagery and the main visitor journey separately, identifying where an asset rather than the prompt supplied the visual advantage.

Review the process as well as the appearance

Luke discusses repeated landing-page styles, workspace organisation and Codex self-checking through headless Chrome. Use those as different review questions: a coherent folder and an attempted browser check are useful process evidence, but neither automatically validates the finished page. The source does not publish the full prompts in its description.

1. Inspect the generated files and any browser-check results available with your own run.
2. Choose a targeted revision for a specific mechanical or visual weakness, preserving the working parts.
3. Record the follow-up work before choosing between a detailed brief and a broad creative request for the next project.

The takeaway

Choose prompt detail for the task’s constraints, then judge actual behaviour and visual output rather than prompt length alone.