# Inference Cost Tracks Context Recycled Not Output Produced ## Claim The price of an AI task is set by how much text has to be held and re-read, not by how much text comes out, which is why a short document can cost more than a long piece of generated code. <!-- atomicity: 1 idea ✓ --> ## Scope & conditions - Holds where input context substantially exceeds output length and must be repeatedly re-supplied to the model across a task (e.g. documentation, consolidation, multi-file readme generation). - Breaks where prompt caching makes re-supplied context cheap, and for pure generation tasks with minimal input context. - Does not establish an exact cost ratio between tiers — no figures are given in the source, only the mechanism. ## Evidence Christopher S. Penn, responding to a viewer who spent over $100 having Claude write five readme files: documentation work is expensive because it requires holding and recycling large text volumes in working memory across the task, whereas code generation typically doesn't. He calls this the "heavy context penalty." > [!quote] > It consumes a lot of working memory to read and write big text files. That's why writing code is relatively inexpensive. ## Links - [[3. Resources/Permanent Notes/Route Judgment to the Expensive Model and Bulk Execution to the Cheap One]] — because extends; that note routes by decision-vs-execution, this one supplies the underlying cost mechanism (context volume) that makes bulk execution cheap to route away from the expensive model. - [[3. Resources/MOCs/AI Cost & Tokenomics MOC]] — because applies-in; this is a general pricing mechanism the MOC's routing guidance should account for. ## Tests / falsifiers - Regress observed cost on input-token volume versus output-token volume across task classes; if output volume dominates the regression, the claim is refuted. - If prompt caching reduces documentation-task cost to parity with code generation at equal output length, the claim holds only for uncached workloads, not as a general property of context volume. ## Maintenance - Last tended: 2026-08-16 - Next review: 2026-11-16 ## Source - [[3. Resources/Literature Notes/You Ask, I Answer - Multi Agent AI Orchestration Costs]] — reports the mechanism and the routing prescription that follows from it.