Build step: one AP task through the whole rhythm
- Status
- stable
- Owner
- Panaversity
- Approved
- Panaversity ·
In this lab you run Dave's reconciliation twice. Run 1 uses his one-line request. Run 2 uses a full first 10 percent that you write. Both get the same four files, including the policy, so either run may succeed. Then you review both outputs as the final 10 percent, before you see the answer key. You turn what you learned into Draft 2 of the AP Worker's Role Contract. Then you port its authority line on paper to each AI vendor's scheduled tasks. It takes about 2 hours of active work. The files come in brightline-lab-ch03.zip, from the Labs companion. When you unzip it, you get one folder with every file the lab needs, including a sample Draft 1 for readers who skipped Chapter 2.
The task. Midwest Packaging's statement says Brightline owes $8,083.50. Brightline's register says $4,083.50. AP policy section 7 says the whole $4,000.00 must be explained, and that the register is never changed to match a vendor. Five items explain it. Some are timing differences and need no action. Others are real and need a person's decision.
Who supplies what. The controller supplies the policy, the deadline and the approval decisions. You write the first 10 percent and do the final 10. The worker does the middle.
What you do. Download the zip, unzip it, and open LAB.md. It gives each step in full, with timings, from Part A to Part I. Follow it in order. You can read it in any text editor. Or upload only LAB.md to Claude or ChatGPT and ask it to guide you through one part at a time. Use a separate conversation for that, not one that the lab asks you to open. In short, you will:
- Predict. Before any run, write down what a one-line request will return and what, if anything, it will get wrong.
- Run. Run 1: paste the one line, attach the inputs, and do not steer. Run 2: copy
briefs/first-ten-template.mdtobriefs/statement-rec-first-ten.mdand fill it with intent, scope, authority and the five review questions. Then run it and stay out of the middle. Record inresults/rhythm-log.mdhow long each part of both runs takes. - Investigate. Review both outputs with
results/review-sheet.md, still without the answer key, and decide to approve, fix, send back or fix the brief. Then score both runs with the rubric. For every point lost, ask the three questions from Concept 3.6 to find the part that broke. - Modify. Write Draft 2 of the Role Contract. Give reconciliations their own authority line: observe and recommend only. Add the stop rule, send corrections and vendor contact to Dave, and add the statement to the evaluations. Add the review checks Dave will use before he approves. Then port the authority line. Write how each AI vendor's scheduled tasks would enforce it. Mark which limits a product setting can enforce, and which only the brief and the review can. This is a paper exercise, so a free plan is enough.
- Make. Apply it to your vertical. Write the first 10 for one recurring task in a role you know, with a stop rule and a "never automatically" line. Also write a five-line checklist for its final 10. Then refine that role's Role Contract the same way: add the checklist as its review checks.
Compare the two runs and the two review times in your log, and report what actually happened, even if Run 1 did well. The lab measures how much each run asked you to trust, and how long you needed to check it.
Artifact checklist
Before you move on to Chapter 4, check that you have finished these. The first five are files in your lab folder. The last one is your own.
-
briefs/statement-rec-first-ten.md, with intent, scope, authority and all five review questions answered -
results/run-1-output.mdandresults/run-2-output.md, both reviewed withresults/review-sheet.mdbefore the answer key was opened -
results/rhythm-log.md, with timings, decisions, scores, and the part that broke for every lost point -
role/ap-worker-role-contract.md, Draft 2, with the reconciliation's authority line, the stop rule, the new evaluation and the review checks -
results/authority-port.md, with each limit set for both AI vendors' scheduled tasks and marked impossible, person decides, or brief and review - A first-10 sheet and a five-line final-10 checklist for one task in your own vertical, and your Role Contract refined with its review checks
3.7 How the two leaders support the rhythm
How the two AI vendors support the rhythm in work that runs on a schedule, the difference that changes a design, and what stays true on both.
Check yourself
Recall and practice for the whole chapter: the flashcards, and a final quiz round from all seven concepts.