活動幀：針對智能體記憶與重播的確定性螢幕活動編譯

摘要

電腦操作代理為了重新推導使用者已經執行過的例行程序，須付出完整的前沿模型推理成本，因為代理的記憶目前記錄的是使用者「說了什麼」，而非使用者「做了什麼」。我們以一條確定性、零模型的管線，將被動捕捉的螢幕活動編譯成代理記憶：該管線將本機捕捉串流切割為具類型的活動幀，以及帶有應用程式、網站、時間、輸入量與指向原始列證據指標的有界事件片段；由於迴路中沒有模型，輸出在位元組層級完全相同、可快取且可機械稽核。在一位專業人士的單一使用者語料庫（51個活躍日，128,756幀）上，編譯器將一天的原始捕捉縮減為可直接作為提示的上下文區塊，縮小86倍，耗時68毫秒；讀取該區塊的代理回答當天相關問題時，對照獨立基準的準確率為98.4%（Wilson 95%信賴區間 91.7–99.7%），而同一捕捉內容交由LLM生成摘要後的準確率僅有66–80%；中階模型讀取該區塊即可匹配前沿模型。同一編譯器同時兼作需求面成本工具。由於其讀取的是委派前被動捕捉的人類活動，而非代理執行軌跡，它能提供代理成本模型假設存在、但據我們所知從未測量過的兩個參數：例行程序開銷比R與例行程序重現率h。我們報告R的首批數值（模型化的上限）為60–343倍，並報告可委派重現率在樣本內為9.0%、樣本外為7.7%，對應於接近8%的實際全代理群token上限；編譯後的例行程序可在模型不介入迴路的情況下被確定性重播，並在一個經守衛比對的命中案例上以零模型token即時示範。結構描述、編譯器與評估工具均已開源。

English

Computer-use agents pay full frontier inference to re-derive routines their user has already performed, because an agent's memory today records what the user said, not what the user did. We compile passively captured screen activity into agent memory with a deterministic, zero-model pipeline: it segments a local capture stream into typed activity frames, bounded episodes carrying application, site, timing, input volume, and evidence pointers back to the raw rows, with no model in the loop, so the output is byte-identical, cacheable, and mechanically auditable. On one professional's single-user corpus of 128,756 frames over 51 active days, the compiler reduces a day of raw capture to a prompt-ready context block 86x smaller in 68 ms, and an agent reading that block answers questions about the day at 98.4% accuracy (Wilson 95% CI 91.7-99.7%) against an independent oracle, versus 66-80% for an LLM summary of the same capture, a mid-tier model reading the block matching a frontier one. The same compiler doubles as a demand-side cost instrument. Read off passive, pre-delegation human activity rather than agent rollouts, it supplies two parameters that agent-cost models assume but, to our knowledge, have not measured: the Routine Overhead Ratio R and the routine recurrence h. We report first values of R, a modeled upper bound, at 60-343x, and a delegable recurrence of 9.0% in-sample and 7.7% out-of-sample, for a realistic all-fleet token ceiling near 8%; a compiled routine replays deterministically with the model out of the loop, demonstrated live at zero model tokens on a guard-matched hit. Schema, compiler, and evaluation harness are open.