Not a dashboard. Not branding. I am running these measurements on myself. The instrument is being built in the open, and this page shows its honest state — every measurement carries one of three statuses. Where a chart is empty, no reading has been published: an empty chart under a signed commitment beats an invented trend.
The live signals — transcript reasoning ratio, role-frames, token volume, provider concentration, and provider list prices — read from a working corpus that already holds 23,865 agent exchanges and 229,453 token-usage events from my day-to-day building. Readings publish signal by signal as the pipeline from that corpus to this page comes online.
I commit to publishing unfavorable trends with the same visibility as favorable ones. Metrics will be locked with a PhD collaborator (engagement pending) before any of them go public. Methodology changes are versioned and visible. The same commitment appears in full on Research, on the page where the questions are named.
This statement is what makes the rest of this page worth reading. Without it, what's above would be a vanity dashboard.
I am the primary subject of this study, by default. Every measurement on this page runs against the working corpus described on Remaining Viable — the same one that feeds Resources.
Mentees will be able to opt in — extending the readings across a cohort and making the data richer. No cohort data is being collected yet: the consent instrument publishes before the first mentee stream does. Consent will be explicit and granular. Mentee identity is preserved. Mentee data never appears on the marketing surfaces of the site.
Opting in is never required to be mentored. The engagement and the inquiry are decoupled by design — no data sharing, no measurement participation, no change to how a mentee works unless they choose it.
Streams a mentee's opt-in will cover: agent transcripts, capability benchmarks, token-cost data, time-allocation data, and — on a separate, optional consent track — the reflection journal underlying Vocational tracking.
An agent transcript is the back-and-forth between me — or the mentee — and the AI tools we use, captured the way a recorded meeting yields a transcript for analysis.
Four artifacts will live here, published before the first quarterly reading does:
What's measured, how, and how often — published as per-dimension protocol documents, versioned.
A PhD collaborator, engagement pending. Will lock metrics before any of them go public and review every quarterly reading.
Every change to methodology, dated and visible. No silent revisions.
The full text of the mentee opt-in. Right to withdraw at any time. Anonymize-in-place or full removal at mentee's choice.