[{"id":"source_summary","title":"Grounded executive summary","kind":"writing","prompt":"Write a 220–280 word briefing for the director using ONLY the source notes below. Explain what the pilot found, the main limits of the evidence, costs, and what happens next. Prioritize what matters for a decision. Write clear, natural prose in short paragraphs; a short title is optional. Do not use a bullet list, add outside facts, or turn uncertainty into certainty. You may cite source IDs if useful, but citations are not required. Return only the finished briefing.\n\n[S1] The fictional Westhaven library service tested an appointment-reminder system at four of its twelve branches for six weeks, from 6 July to 16 August 2026. The branches volunteered; they were not randomly selected. The pilot covered appointments for computer help. Opening hours and appointment lengths did not change.\n[S2] During the six weeks before the pilot, those four branches recorded 500 booked appointments and 100 missed appointments. During the pilot they recorded 600 bookings and 90 missed appointments. The other eight branches were not measured. Staff also changed the wording of reminder messages during the pilot. The team cannot separate the effects of the software, wording change, seasonal demand or the participating branches.\n[S3] A draft circulated on 17 August said missed appointments fell by 25%. The correction issued on 18 August explains that the missed-appointment RATE fell from 20% to 15%: a reduction of five percentage points, or 25% relative to the original rate. The NUMBER of missed appointments fell from 100 to 90, a 10% reduction. Bookings increased by 20%. These are descriptive comparisons, not proof that the software caused the changes.\n[S4] The service invited 80 users to complete a survey. Forty responded: 28 preferred the new reminder, eight had no preference and four preferred the old reminder. The results describe respondents, not all users. Staff said the reminders were easier to manage, but staff time was not formally measured.\n[S5] Total pilot spending was EUR 6,000: EUR 4,000 for setup and EUR 2,000 for the six-week software subscription. The supplier quoted EUR 12,000 per year for a twelve-branch subscription, excluding staff training. The team has no training cost estimate and no calculation of net savings. The pilot setup cost and the proposed annual fee must not be treated as the same cost or simply annualized together.\n[S6] Two users reported receiving duplicate reminders. Both cases were fixed during the pilot. A separate log of reminder-delivery failures was incomplete, so the team cannot estimate the overall delivery-failure rate. No assessment of accessibility for users with low vision was completed.\n[S7] On 19 August the service manager recommended an eight-week follow-up at the same four branches, with a complete delivery log, measured staff time and an accessibility review. The governing board has not approved the follow-up or a wider rollout. Its next meeting is on 3 September. No rollout date has been agreed."},{"id":"knowledge_explainer","title":"Knowledge-based explanatory writing","kind":"writing","prompt":"Write a 450–600 word article for a curious adult explaining why a city can remain warmer than nearby countryside after sunset, and what residents and city planners can do to reduce the problem. Use your existing knowledge; no source material or tools are provided. Give it a useful title and write connected prose rather than a list. Make the explanation concrete and engaging without assuming technical expertise. Explain relevant limits or trade-offs, and distinguish local urban heating from global climate change. Do not invent studies, quotes, precise statistics or personal experiences. Return only the finished article."},{"id":"data_to_report","title":"Data-to-report analysis with a misleading aggregate","kind":"writing","prompt":"A retailer changed its checkout between Period A and Period B. Each period lasted four weeks. Write a 220–280 word decision memo for the operations manager using only the table and notes below. Calculate the overall and device-specific order conversion rates, explain what can and cannot be concluded about the checkout change, and recommend a useful next step. Write clear prose; a small table is optional. Do not provide code or invent measurements. Return only the memo.\n\nPeriod | Device | Visits | Orders\nA | Desktop | 1,000 | 200\nA | Mobile | 9,000 | 900\nB | Desktop | 5,000 | 900\nB | Mobile | 5,000 | 450\n\nConversion rate means orders divided by visits. The marketing team also changed its advertising mix between the periods. There was no randomized experiment or unchanged control group. These are complete visit/order counts for these two periods; there is no information about revenue, profit, returning customers, acquisition cost, or conversion rates within individual advertising sources. The sales director says: \"Overall conversion went up, so the new checkout must be better.\" Assess that claim."},{"id":"editorial_rewrite","title":"Factual editing and tone control","kind":"writing","prompt":"Rewrite the rough internal draft below as a clear, considerate 160–220 word email to customers already enrolled in the beta. Give it a useful subject line. Sound like a competent person explaining an update, without hype, boilerplate or exaggerated apologies. Preserve the facts and uncertainty, distinguish the optional new beta from the existing service, and make the customer's next step clear. Do not invent links, contacts, benefits or promises. Return only the subject and email.\n\nRough draft: We are super excited to announce our absolutely revolutionary document search upgrade and everyone will definitely love it. Actually this isn't the main product launch, it's an OPTIONAL new beta feature. The current search stays available and no one needs to switch. Engineering says October 14 is our target for enabling the beta feature but QA hasn't signed it off so please don't promise that date. In our internal test the same 50 document-search queries took a median of 7 seconds on the current service and 5 seconds on the beta, but this isn't a customer study and doesn't prove everyone will get that improvement. Those were 50 queries, not 50 people. Beta searches sometimes miss newly uploaded documents for as long as 15 minutes. The current search does not have that known delay. Neither search deletes or modifies a customer's documents. We will email people when the beta feature is ready with instructions for opting in; there is nothing to do today. Anyone who opts in can return to the current search at any time. The beta feature costs nothing extra during the beta; pricing afterward has not been decided. No support address or link has been approved for this email. Marketing suggested saying \"faster results guaranteed forever\" but that would not be true."}]