Experiment 002 · Writing
Begin a sentence. We'll finish it.
Pick an opening, then a continuation. A rule-based sequence model ranks where the sentence is most likely to go — and shows the probabilities out loud.
Pick an opening that calls to you.
Each opening has a different shape of likely completions.
Read the opening, then pick what feels right.
Reveal
Ranked likelihood
Least likely (our model)
Most likely
- Your pick
- —
- Our top prediction
- —
- Match?
- —
- Your pick's rank
- —
Why these rankings
What the model doesn't know: your dialect, mood, or what you read this morning. The ranking is a small, auditable set of hand-tuned priors over four openings — not an LLM. The goal is to make the mechanics visible, not to maximize fluency.