Lesson 6 · Functional behaviorCourse indexPT-BR
CLIProxyAPI benchmark · Functional behavior

Measure each lane's real continuation

Resending, remote state, local resume, and compaction are different forms of memory.

Primary local sources: BENCHMARK-MEASUREMENT-PLAN.md, model-context-matrix.csv, and planned-run-manifest.json.
BENCHMARK-MEASUREMENT-PLAN.md · model-context-matrix.csv · planned-run-manifest.json

The idea in plain language

Direct Chat and Messages resend history. Responses uses previous_response_id when truly supported. Warp measures its real session and, separately, sidecar emulation. Codex uses codex exec resume; Claude Code uses --resume or --continue.

Running example: send the same sealed parcel — nonce, checksum, and three needles — through all six lanes. If the harness changes the wrapping, you are measuring the lane too. The analogy breaks because context, tools, and thinking carry semantics, not size alone.

Three turns recover a nonce and checksum; the tap proves how many tokens were sent. Compaction, a volatile map, local state, and full resending get distinct labels. The benchmark does not promote one mechanism to another just because the final answer looks correct.

Verifiable excerpt

direct_chat/messages = explicit_history\ndirect_responses = previous_response_id when proved\ncodex = exec resume · claude = --resume/--continue

outputs/BENCHMARK-MEASUREMENT-PLAN.md · outputs/model-context-matrix.csv · outputs/planned-run-manifest.json

The path in one picture

UI / CLIobserved outputharnessWarp · Codex · Claudecompatibilitysidecar / envelopeCLIProxyAPI127.0.0.1:8317providerdirect Chat · Responses · MessagesPRIMARY BOUNDARY
Running example: the same nonce + checksum crosses every lane; the failure belongs to the smallest boundary where it reproduces.

Retrieval check

What must the tap prove during continuation?