Every AI picks 7: Anthropic engineer's rule for escaping model taste convergence
rubenhassid · x · 2026-10-16
Ask humans to pick a number between 1 and 10 and about 1 in 3 says 7. Ask Claude, ChatGPT, Gemini, or Grok—it's always 7. People are biased; AI is far more biased. When everyone bases decisions on the same "7" ("Is this post good?" "Which logo?" "What should I name my product?"), taste disappears.
The rule: never ask AI for its opinion—use it to find the 7, then go somewhere else. The insight comes from Anthropic engineer Felix Rieseberg's essay, which quotes Terence Tao: models are "astounding in some directions, while inadequate in others." His core rule: never let the model judge its own work.
Screenshots show the technique in action: forcing a model to estimate the probability another AI would produce each tagline and requiring the last two under 10% to force differentiation—and four models giving the exact same science joke.
More from Models
- Unverified claim: Apple and Meta's 816B-parameter Muse Spark Open runs on two Mac Studios — TansuYegen · 2026-10-16
- Gemini 4 model card: model eliminated months of trial-and-error on bioweapons research — Puzzleheaded-King584 · 2026-10-16
- Brain teaser: strip all attention, MLP and embedding matrices from a Llama 3 — how many trainable params remain? — Sauers_ · 2026-10-16
- Lean v4.35.0 ships verified kernels — one written and proven consistent by Claude — geoffreyirving · 2026-10-16
- One flagged seat froze an entire Claude team org — founders share backup plans — RexDouglass · 2026-10-16
- Kimi K3 checkpoint impresses: strong held-out results point to solid pretraining — SonglinYang4 · 2026-10-16