The Working Set of a Coding Agent: Coherence Debt in Repository-Scale Tasks

2026-08-17

AI Summary

This study addresses factual omission and inconsistency in warehouse-scale coding agents caused by limited context windows by modeling the task as coupled fact graph reconstruction and introducing the concept of "coherence debt." Through multi-model fault injection and SWE-bench evaluation, we demonstrate that fact availability is a more critical determinant of task success than code proximity, with missing facts leading to erroneous generation rather than safe termination. Furthermore, this work reveals an order-of-magnitude variance in token consumption across different toolchains and establishes that read operations no longer effectively predict success rates. These findings provide a theoretical foundation and a novel paradigm for optimizing agent context management strategies in large-scale software engineering environments.

Abstract

Repository-scale coding requires an agent to keep tests, imports, configuration, and migration rules consistent within a bounded context window. We model this as reconstructing a coupled-fact graph: at each edit, a required fact comes from recent context or parametric memory, and the facts covered by neither form coherence debt. We supply and withhold each channel and inject faults across seven models and five harnesses. As expected, no model completes a task on an unseen API with both channels empty, and putting the facts in the prompt restores success. When a rename defeats what models memorized about a real library, all seven fail in the same place, passing and missing the same tests. Availability decides the outcome and distance does not: withholding a fact costs exactly the work it supports, and a supplied fact works as well far from the edit as next to it. Harnesses pay unequal prices for it: configurations that all pass every test differ more than tenfold in tokens consumed because they rebuild the same content at different rates, and spending more recovers nothing when facts are withheld. A missing fact produces wrong work rather than absent work: an agent asked to act acts, fabricating the file or guessing the value, so instruments built on reads look for a hole already filled. How often it says it is blocked instead is a property of the model, from every trial to none. Availability does not settle every edit: where standard and code disagree, agents follow the standard even when it prescribes the worse code, so a stale convention file costs more than no file. Because parametric memory substitutes for reading, on SWE-bench, where models likely know the repositories, reads no longer predict success. Harnesses should keep the facts an edit depends on available when the agent writes, and check that availability against what the agent produces rather than what it reads.

Problem

• Coherence Debt
• Repository-Scale Coding
• Context Window
• Coding Agent
• Fact Availability

Innovation

• Coherence Debt
• Coupled-Fact Graph
• Repository-Scale Coding
• Parametric Memory
• Fact Availability