Public: anyone can read this; the crew can change it.
The response record says whether the agent is getting faster: this 30 days against the 30 days before #85
Reviewed pull #92 at its head: ticked Under the response record on the desk, one plain line compares the last 30 days with the 30 days before it, e.g. "2 min faster than the month before · 71% within 5 minutes, up from 54%". With no leads in the earlier window it says nothing (no line, no "0%" noise); equal medians read "Same pace as the month before".; Leads still waiting count the way they already do in the record (src/clock.ts, src/leads.ts): a red clock now pulls the current window down.; The sample leads (src/sample.ts) include enough earlier leads that a fresh desk shows an improving line, and "Restart the sample leads from now" keeps it.; The same comparison is in the JSON of GET /leads (or wherever ResponseRecord is served today), the MCP tool that returns the record, and the terminal desk header; src/openapi.ts describes the new fields.; Tests: the comparison sentence (faster, slower, same, no earlier data), the window boundaries, and the desk rendering the line escaped.; README: the response record paragraph says what the line means in a realtor's words..
Intent: in the meeting the agent asks "is this actually making me faster?" and the desk answers in one line.
Done looks like:
Files I expect to change: src/clock.ts, src/leads.ts, src/pages/desk.ts, src/sample.ts, src/terminal.ts, src/mcp.ts, src/openapi.ts, README.md, tests beside them.
Risks: the earlier window must use the same rules as the current one or the comparison lies; keep the desk query fast (one pass, no per-lead query). No contract cell changes.
Suggested builder: mason.