Brief: Position and proportion set what a reader retains

Type: types/write-brief.md

Intent

Explain how text position can influence perceived importance and recall, for human and LLM readers separately. Present the proposed effect of relative space allocation as a conjecture, not as an established result of position studies. The reader should understand why an intelligible opening claim can guide human theme identification, why support that outgrows its role might compete with it, and why theme identification, recall and LLM judgment must remain separate outcomes. Five criteria currently assume this without stating it: core-claim-obscured, proportion-mismatch, detail-overhang, branch-bloat, and the surface-the-divergence move in the readability-and-flow pass, together with the Churchill cue and the operator-brief rule to start with what happens.

Must keep

• Two readers, two evidence bases. The LLM half may rest on the KB's position-bias ingest and on LLM context is composed without scoping; the human half needs at least one ingested source on first-mention or serial-position effects in text comprehension. Do not let one reader's evidence stand for the other's.
• Proportion is relative to explanatory role, not an absolute length or a requirement that support be shorter than the claim. Explain the conjecture that disproportionate support competes with the claim, name a test, and preserve necessary support behind a pointer when that improves the reading path without hiding conditions needed to understand the claim.
• The boundary with title as claim enables traversal as reasoning: that note covers position at index level; this note covers position inside a document.
• The brevity hazard: shortening alone need not repair a misplaced claim. Moving the claim addresses placement directly. Do not imply that deleting preceding material can never help; the repair must address the actual cause.

Must exclude

• A word budget, section budget or any numeric threshold. The compression bundle criteria already carry the operational tests.
• A restatement of the compression bundle's four criteria. Link to them as the procedures that apply the mechanism.
• Claims about recency or middle-position effects in LLMs beyond what the cited ingests support.

Reserved decisions

• Use Kieras (1980) and Freebody and Anderson's 1981 Technical Report No. 219 for the human side. Keep one note with distinct human and LLM evidence.
• The operator approved treating proportion as a conjecture with a test; no position finding may be presented as a direct test of proportion.
• Do not expand the LLM judgment-order result into a result about retention or comprehension within one document.