Generate, schedule and review reports

1. Before generation

1. Verify management permission for the grant or project. Platform administrators, PIs and lead researchers generate within their managed scope.
2. Inspect the latest collected bodies, source dates and researcher assignments in Materials. AI does not independently discover every uncollected external document.
3. An administrator checks AI enablement and runner readiness in Lab management → AI/Drive settings. Review your allowance under Profile → AI usage.
4. Enter plans, objectives and desired reporting perspective under Research goals/AI context. These are reference context, not model retraining.

2. Generate a report manually

1. Open Reports → New report → Draft with AI.
2. Type and select grants, projects and researchers from the dropdown. Add multiple items and remove with ×. Review the scope to avoid unrelated evidence.
3. Choose daily, weekly, monthly or a custom period and check the start/end dates when shown.
4. Submit, then follow Pending, Running, Ready or Failed in the report list. Avoid repeated submissions while waiting. The same target and period may reuse an existing report; choose a new version only when regeneration from updated evidence is intended.

3. Create or edit a schedule

1. Open Activity reports → Schedules → Create schedule.
2. Choose a lab and target. Recurring schedules currently accept one grant, project and researcher each. Use manual generation for a one-off combined target.
3. Choose the period and execution hour: 8 means 08:00 and 20 means 20:00, Seoul time. Daily covers yesterday; weekly runs Monday for the previous Monday–Sunday; monthly runs on day 1 for last month.
4. Check Enable this schedule and Save schedule. Verify Active and Next run on the card. Use Edit schedule on that card to change its settings.

4. Disable a schedule clearly

1. Choose Pause schedule to stop temporarily, or Cancel and delete to remove the schedule; no other fields need to be filled.
2. For pausing, verify Paused and No next run. Unchecking Enable this schedule in the editor and saving has the same effect.
3. Pending requests belonging to that schedule are cancelled. Running or completed reports are not stopped or deleted.
4. Paused schedules stay in the list with Resume and Edit controls. Cancel and delete removes the schedule itself while retaining existing reports.

5. Review, annotate or delete a result

1. Open a completed report and check its subtitle targets and reporting dates.
2. Follow numbered evidence links to expand source material. Distinguish dates, actual work and decisions from AI interpretation. Missing evidence does not establish that no work happened.
3. Edit metadata to save tags, notes and assigned researchers. Add multiple researchers by typing and selecting. Assignments do not recalculate the original generation scope.
4. Confirm deletion to move an unwanted report to Trash. Disabling a schedule and deleting a report are separate; Drive copies and operational backups are not removed.

6. Choose instructions, privacy and completion email

1. Select and load an instruction template, then add the questions for this report. Follow Reusable report instructions for details.
2. Select Only me for private review and the full-body email option if needed. Delivery uses the author's profile email.
3. Review multi-scope targets and dates for immediate generation. Schedules use their saved targets, time and instructions on later runs.
4. To write the body yourself, use Write report rather than AI generation. See the direct-writing/privacy/email chapter.

7. Read the common sections and handle failures

1. New ordinary AI reports use Summary → Progress against plans → GitHub development → Researchers/collaboration → Needs verification → Paper preparation → Proposed next steps. Magazine articles use an editorial format.
2. Open sources to distinguish observations, interpretation, proposals and undecided matters. Missing evidence may leave a section limited instead of generating invented content.
3. If a run failed, inspect the same report first. Do not submit again if a requested retry has already completed.
4. For persistent queues/failure, check collection, scope, period, AI enablement and allowance. Send Support the report URL and request time. A Notion publishing failure is separate from AI report generation.

Write AI and manual magazine articles

1. At the bottom of the magazine home, use Write an AI article or Write an article manually. These compact links are available to lab, grant or project managers.
2. Select the lab, related grants/projects/researchers and evidence dates. Scoped managers must select a scope they manage. Multiple selections are supported.
3. AI: enter a topic and instructions describing the questions, process, readers and structure. Generate the article and refresh its status. Enabled AI, available quota and accessible collected evidence in the date range are required. Narrow the scope if too many sources are selected. Finished articles have a whole-article summary and source list.
4. Manual: enter the title, citation-free summary and Markdown body. Use heading, bold, table and preview controls, @mentions and [[document search. Save a draft for yourself, continue through Edit article and publish for the selected audience. Authors can edit their published articles.
5. Editorial articles are separate from researcher awards. Related researchers do not automatically receive badges or prizes. The archive distinguishes AI and manual articles.

Find reports and reproducibility results