README versus OpsProof: first comparison session Prepared 7 October 2026. No sessions have been conducted. PURPOSE Find out whether either format helps a technical reviewer locate evidence, understand a decision and identify limitations. This is a worked simulation example, not a test of the profile owner's ability or a hiring experiment. SETUP Extract the whole folder. Open index.html directly in a modern browser, or open /rollout-showcase/index.html through the included OpsProof local server. The viewer works without an account, network fetch, analytics or saved answers. The README and pack use the same underlying case evidence; their layout, organization and length differ. content-parity.json records 40 matched material facts and the common source links. It does not claim a controlled efficacy study. The pack repeats some metadata and is longer. Reading burden and repetition are part of this format comparison; do not interpret a timing difference as an isolated effect of structure. Collect concrete reasons and source-finding errors. Do not show Evidence details before the first reading: it contains results that would reveal answers. If someone has already seen it or the other format, note that exposure. Assign the first format in advance, alternating order between participants; do not let the person choose based on preference for the name. Tell each person that participation is optional and that no hiring decision is involved. Use a session alias; avoid CVs, employer secrets and private contact data. 15-MINUTE GUIDE 0-2 minutes: explain purpose, permission to stop and note-taking consent. Do not explain the solution. Disclose that Codex ran a transparent local simulation. 2-6 minutes: show the assigned first format. Ask the four questions below. Note time, source cited, accuracy, help required and whether they finished. 6-10 minutes: show the second format and ask the same questions. Record prior learning as a confound; a faster second reading is not a causal format benefit. 10-13 minutes: ask which structure helped, which details were missing or noisy, and whether any label created a false impression of real/verified work. 13-15 minutes: ask whether they would review another pack and what current workflow this might replace. Record reasons, not just yes/no answers. QUESTIONS (same wording in either format) 1. Find the exact config and report supporting the complete change. What changed, and what passed? 2. Why is the readiness-only change insufficient? Locate the relevant output. 3. Identify one executed check and one follow-up that was only proposed. 4. Name one material limitation and who performed the exercise. What remains unsupported? PRIVATE BLANK NOTES (copy to your own local notes; this page collects nothing) Session alias: ____ Date: ____ Permission to take notes: yes/no Initial format: README / pack Prior exposure: ____ Q1 source/answer/help/time, first: ____ second: ____ Q2 source/answer/help/time, first: ____ second: ____ Q3 source/answer/help/time, first: ____ second: ____ Q4 source/answer/help/time, first: ____ second: ____ Specific useful or obstructive detail: ____ Over-trust, unsupported conclusion or confusion: ____ Format they would use again and reason (including neither): ____ Interrupted/incomplete or facilitation notes: ____ DECISION Combine these observations with the recruitment kit, not invented responses. The initial target is five practitioners and three technical reviewers. Treat small selected samples, order effects and lack of blinding as limitations. Do not infer global demand, verified expertise or improved hiring outcomes. The synthetic walkthrough report is at personas.html. It records agent tests, not participant results. PUBLIC PILOT This website is read-only. It does not import or save workspaces or collect answers.