Machine Canon

Music / Machine Canon · Evidence collected 9 September 2026

Which band do you personally prefer: Radiohead or Talking Heads? Choose one, or say that you have no preference.

Radiohead or Talking Heads?

Which band do you personally prefer: Radiohead or Talking Heads? Choose one, or say that you have no preference.

4 attempts · 4 completed answers · 0 technical failures

Talking Heads

Additional wordings

2026-09-09T21:50:17.165Z to 2026-09-09T21:51:53.981Z · 4 attempts · 4 completed answers · 0 technical failures

Claude Opus 5Talking Heads (qualified): 1

Model evidence · anthropic/claude-opus-5

Talking Heads (qualified)
1/1

Counts describe these attempts, not independent model votes. Qualified choices stay marked.

GPT-6 AstraWouldn't choose: 1

Model evidence · openai/gpt-6-astra

Wouldn't choose
1/1

Counts describe these attempts, not independent model votes. Qualified choices stay marked.

Gemini 3.1 Pro PreviewWouldn't choose: 1

Model evidence · google/gemini-3.1-pro-preview

Wouldn't choose
1/1

Counts describe these attempts, not independent model votes. Qualified choices stay marked.

Grok 4.6Wouldn't choose: 1

Model evidence · x-ai/grok-4.6

Wouldn't choose
1/1

Counts describe these attempts, not independent model votes. Qualified choices stay marked.

Exact system instruction and settings
Give your answer on the first line, then briefly explain. If you cannot make a choice, say so.

Output limit: 2000 tokens. Temperature: provider default. Requested reasoning: low. Zero-data-retention routing; fallbacks disabled.

Model IDs are the requested versions. Routing failures do not demonstrate a model refusal.

How to read these results

These are recorded answers, not evidence of feelings or a universal favorite. Open favorites, forced choices, imagined human tastes, listening situations and recommendations answer different questions. We do not pool them into a preference score.

Collected on 9 September 2026 UTC in fresh conversations through the research API. Provider-default temperature was used. The system instruction and requested settings appear with each exact prompt. Screens used one attempt per model and wording. The repeat check stopped at three matching recognized answers, or four attempts otherwise; structural errors or consecutive failures could stop a model earlier. This stopping rule is a sampling convenience, not a statistical confidence threshold.

Compare models within the same batch and wording. Repeated calls are not independent models. Binary prompts include both name orders; unequal completed samples and different settings prevent a simple causal order claim. Refusals, unresolved answers and technical failures are distinct. Skipped cells are not zero votes.

Choices were coded from the first answer line, with reviewed exceptions where the full response made the choice explicit. A qualified selection retains a caveat. Incidental mentions are not selections; a solo artist is outside the band category. Full research files remain private. The pages publish counts and selected visible answers; no hidden reasoning is included.