Direct answer · the boundary problem
Stop making the next session reconstruct the project — hand it a record instead. Compaction and session death are where AI coding work quietly loses its memory. The summary that survives is lossy and confident; the touched files can be expensive to re-read. A compact, hash-chained work record can be smaller to read than those files, and it can be checked.
Two example payloads from a real multi-file task (RatePilot team seats), both counted with tiktoken o200k_base. Actual token savings have not been measured. Method, payloads, and limits.
When the context window fills, the session gets summarized and work continues from the summary. That summary is lossy by construction: file contents, decisions, and verification state compress into prose, and the next stretch of work proceeds confidently from whatever survived. Nothing tells you which details were dropped — the model sounds exactly the same either way.
Keep durable state outside the context window. Even without special tooling, three habits help:
The structural version of those habits is VeriCommand: work is recorded as packets, dispatches, signals, and returns on a hash-chained record, and the resuming session reads that one compact record. In one example from a real task, that read is 1,013 tokens, and re-reading the eight touched files is 22,897, both counted with tiktoken o200k_base. Actual token savings have not been measured.
Source: the VeriCommand token benchmark →Because it is confident even where it's wrong, and there's no way to check it from inside the session. In one recorded production incident, a resumed session's carried summary asserted the update feed was on one version; the board's hash-chained record said another. The record was right, the memory was stale — and the deploy went out correct only because the session was required to read the record instead of trusting itself. A summary is a recollection; a record is evidence.
Using VeriCommand with Claude Code →Yes — the record is model-neutral. Claude Code, Codex, Cursor, or Kimi can each read the same hash-chained record at a boundary. A hand-off between vendors can use the same compact read as a resume within one tool, instead of a fresh reconstruction per model. That's the point of putting the memory in the record instead of in any one agent.
How cross-model hand-offs work →A single unbroken session gets no benefit and pays a small overhead. With no boundary there is nothing to resume, and the governance calls still run: about 427 tokens in a representative example session. The record can only help at a compaction, a session death, or a model hand-off.
See the one-session case →The local core is free. Put the memory in a hash-chained record, and let the next session — or the next model — pick the work back up from one compact read instead of a full re-read.