Where a fact appears can change the answer
Move the same decisive fact through a long document. See why position can matter more than size.
The question is always the same: “Can the buyer cancel before the renewal date?” Only 18 words decide the answer.
Stronger attention Fainter in the middle Stronger attention
Where should the fact sit?
Move the same 18 words. The document and question stay unchanged.
Authored sample · not a live runContext is attention, not storage
A context window is the text available to a model for one task. Fitting a fact inside that window does not guarantee that the model will use it well.
Read Context is attentionFor the curious
This preview uses three fixed outcomes to make one documented evaluation risk tangible. It is not a benchmark, and it does not claim that every model will behave this way on every document.
How this was built
Approach: one synthetic contract, one fixed question and three authored outcomes. Only the position of the decisive fact changes.
Runtime model: none. Inference cost: £0; hosting is separate.
Limit: this is an attention exercise, not measured model performance. Arun will add the personal build story after review.
Build diary awaiting Arun’s notes. Empty fields below are intentional; no build story or time estimate has been invented.
- Tools used
- OpenAI Codex, Next.js 16, TypeScript, Impeccable
- Models used to build
- Arun to add
- Rough hours
- Not recorded
- What failed
- Arun to add
Actual build prompts
Yes build them into the site