# Weekly Sprint Report Axdale. Software engineering by Michael Lip, Warsaw, Poland (CET). Website [https://www.axdale.com](https://www.axdale.com). Code portfolio on GitHub at [https://github.com/theluckystrike](https://github.com/theluckystrike). Version 1.0. 23 October 2026. You get this report by 18:00 CET on the last Working Day of each Week. Part A is the blank template. Part B is a filled example for a fictional client, and it follows the Statement of Work example in the same document set. ## Part A. Template ### 1. Reference | Field | Details | |---|---| | Client | ____ | | SOW number | ____ | | Week | Free Week (week 1) / Paid Week number ____ | | Working Days | ____ to ____ | | Hours used | ____ of 40 | | Report date | ____ | ### 2. Summary Three sentences at most. What shipped, what it means for your users and what is next. ____ ### 3. Acceptance criteria | No. | Criterion in short | Status | Evidence | |---|---|---|---| | AC1 | ____ | Met / Not met / Partly met | ____ | | AC2 | ____ | ____ | ____ | | AC3 | ____ | ____ | ____ | ### 4. Pull requests | PR | What it does | State | Criterion | |---|---|---|---| | ____ | ____ | Merged / In review / Draft | ____ | ### 5. Measurements | Measure | Last Week | This Week | Target | |---|---|---|---| | ____ | ____ | ____ | ____ | ### 6. Hours | Day | Hours | Main work | |---|---|---| | ____ | ____ | ____ | | Total | ____ of 40 | | ### 7. Blocks and time lost ____ ### 8. Production actions | Action | Approved by and when | Result | |---|---|---| | None / ____ | ____ | ____ | ### 9. AI providers used ____ ### 10. Risks | Risk | Likelihood | Effect | What I do about it | |---|---|---|---| | ____ | Low / Medium / High | ____ | ____ | ### 11. Decisions needed from you 1. ____ ### 12. Plan for next Week 1. ____ 2. ____ Next Week estimate ____ hours at the agreed hourly rate. Invoice sent on ____. ## Part B. Filled example (fictional) The client, people and numbers below are fictional. They show how a real report reads. ### 1. Reference | Field | Details | |---|---| | Client | Example Client Ltd | | SOW number | EXAMPLE-02 | | Week | Paid Week 2 | | Working Days | 9, 10, 12, 13 and 16 November 2026 (11 November is a public holiday in Poland) | | Hours used | 38.5 of 40 | | Report date | Monday 16 November 2026 | ### 2. Summary The reply drafter now runs against 40 real past tickets and passes the rubric on 36. The evaluation is a required CI check, so a change that lowers quality can't merge. Next Week adds staff feedback buttons and a staged production rollout behind the flag, with your written approval for each deploy. ### 3. Acceptance criteria | No. | Criterion in short | Status | Evidence | |---|---|---|---| | AC1 | Evaluation set grown to 40 anonymized cases | Met | PR #41, file `evals/drafts/cases.jsonl` | | AC2 | At least 34 of 40 cases pass the rubric | Met, 36 of 40 | CI run on PR #44, report attached | | AC3 | Evaluation is a required CI check on the main branch | Met | Branch protection setting, PR #43 | | AC4 | Each request stops at 4,000 output tokens and the daily spend stops at USD 20 on staging | Met | Tests `drafts.caps.spec.ts`, staging log | | AC5 | p95 latency under 15 seconds on the evaluation set | Partly met, 16.2 seconds | See risk 1 and decision 1 | ### 4. Pull requests | PR | What it does | State | Criterion | |---|---|---|---| | #41 | Grows the evaluation set from 15 to 40 cases | Merged | AC1 | | #42 | Fixes timeouts on tickets longer than 3,000 words | Merged | AC5 | | #43 | Adds the evaluation as a required CI check | Merged | AC3 | | #44 | Rubric update and report output in the CI summary | Merged | AC2 | | #45 | Token cap per request and daily spend cap | In review | AC4 | ### 5. Measurements | Measure | Week 1 | This Week | Target | |---|---|---|---| | Evaluation cases | 15 | 40 | 40 | | Cases that pass the rubric | 12 of 15 (80%) | 36 of 40 (90%) | 85% or more | | Drafts with at least 1 help article link | 14 of 15 | 39 of 40 | All | | p95 latency | 13.8 seconds | 16.2 seconds | Under 15 seconds | | Model cost per draft on staging | USD 0.021 | USD 0.017 | Under USD 0.03 | | Prompt injection cases blocked | 3 of 3 | 6 of 6 | All | ### 6. Hours | Day | Hours | Main work | |---|---|---| | Mon 9 Nov | 8.0 | Selected and anonymized 25 new cases with the support lead's export | | Tue 10 Nov | 7.5 | PR #41 and PR #42 | | Thu 12 Nov | 8.0 | PR #43, CI check and branch protection | | Fri 13 Nov | 7.5 | PR #44, rubric review | | Mon 16 Nov | 7.5 | PR #45, latency tests, this report | | Total | 38.5 of 40 | | ### 7. Blocks and time lost The staging test key hit its spend cap on Tuesday at 14:10 CET. Jane Doe raised the cap at 15:30 CET, and I worked on PR #42 in the meantime, so I lost no time. ### 8. Production actions | Action | Approved by and when | Result | |---|---|---| | None | Not applicable | Not applicable | ### 9. AI providers used Anthropic only, as approved in the Schedule. No other provider saw the Client's code or tickets. ### 10. Risks | Risk | Likelihood | Effect | What I do about it | |---|---|---|---| | p95 latency is 16.2 seconds, above the 15 second target | High | Staff wait longer on long tickets | Stream the draft so the first words show in about 2 seconds, or allow 20 seconds for tickets over 3,000 words (decision 1) | | Help articles older than 12 months give outdated answers | Medium | Some drafts cite old steps | Add the article date to the draft and flag articles older than 12 months | ### 11. Decisions needed from you 1. Choose one. Stream the draft (my default, about 6 hours of work), or accept 20 seconds for tickets over 3,000 words. 2. Name 2 support staff who will try the drafts on staging next Week and leave written feedback in the tool. ### 12. Plan for next Week 1. Staff feedback buttons on each draft, stored on staging. 2. Streaming of drafts, if you choose it in decision 1. 3. A cost and quality page for the support lead. 4. A staged production rollout plan behind the flag. Each deploy needs your written approval. Next Week estimate 40 hours at the agreed hourly rate. Invoice sent on 16 November 2026. --- Axdale. Michael Lip, independent engineer, Warsaw, Poland (CET). michael@axdale.com.