KEEPING UP WITH THE CODE: FILE CONTEXT SYNCHRONIZATION FOR CODING AGENTS

Coding agents change a repository while working, but their prompts usually keep every earlier file read. The prompt therefore accumulates duplicate code and code made stale by later edits. Existing compression methods ask which history is useful; they do not check whether that history is still true. We introduce LiveState, a training-free context layer that keeps file observations synchronized with the repository. It replaces old read outputs with stable pointers to a bounded panel containing only the regions the agent inspected. After an edit, LiveState finds each region in the current file and refreshes its text; if the match is uncertain, it hides the region rather than show stale code. On SWE-bench Verified, LiveState reduces mean turns for all eight models from three providers, while accuracy and token use vary by model. Claude 5 Fable matches baseline pass@1 ($+0.1$ points) with $17.0\%$ fewer turns, and Gemini 3 Flash gains $6.7$ points (paired 95% CI $[+4.3,+9.2]$). The other Claude models trade $1.4$–$3.3$ points for $24$–$35\%$ fewer turns and $23$–$48\%$ fewer tokens. On the longer-horizon DeepSWE benchmark, Fable keeps pass@1 within $0.4$ points while using $37.7\%$ fewer turns and $33.3\%$ fewer tokens, and within an 8M-token budget it resolves $43.4\%$ of tasks versus $24.5\%$ for the baseline. Keeping file context current is therefore not just prompt shortening: it also changes how agents act.