A useful end-of-day review should tell you what still needs attention tomorrow. OGAD (Off Grid AI Desktop) can prepare that review automatically through a God routine, using saved tasks, retained work context, and the connected sources you choose.
Set the routine after you enable the relevant Pro features and source access. Keep the computer and app available at the scheduled time. Automatic review does not mean that every unfinished task in your life has been captured.
Download OGAD | Desktop beta 122 release notes
What would you like to do with Off Grid AI?
Have a feature or use case you would like us to support? Tell us what you want to do and which device you use.
Write to support@getoffgridai.co, join our Slack community, or talk to us on Reddit.
Separate three kinds of unfinished work
Use three groups: work you can continue, work waiting on someone else, and work whose status is unclear. This prevents a long list from treating every open item as something you should do tomorrow morning.
A task discussed during a meeting may never have been accepted. An assigned action may already be complete in a later source. Ask the assistant to distinguish those cases instead of treating every task-like sentence as a current obligation.
Start with one project or your own clearly assigned work. Broaden the review only after the smaller version produces a list you can verify.
Prepare the evidence the routine will use
Install OGAD, enable Pro, and open God with a local tool-capable model. Review your Actions list and the notes or accounts needed for the task.
If you want retained activity included, enable the relevant capture and check its visible state. Sampled screen activity is not a complete recording of all work. It cannot establish what happened in an excluded app, on another device, or before capture was enabled.
Connect live accounts only if the review needs them. A local model can use retained local evidence after setup, but current email and online project information require network access.
Write a review instruction you can audit
Use an instruction like:
Review my Cedar work at the end of today. Use available open actions and today's relevant retained notes and messages. Group results into I can continue, waiting on someone else, and status unclear. Include the latest supporting source and date. Check for later completion evidence. Suggest at most three starting points for tomorrow. Do not mark tasks complete, send messages, or change external records.
If you have a cutoff time, include it. Ask the assistant to state the period and sources it actually checked. An empty source should produce a visible gap, not an invented account of the day.
The three suggested starting points are recommendations for your review. The assistant should not treat them as confirmed priorities without the constraints you provide.
Add the end-of-day schedule
Open God settings and Scheduled tasks, then choose Add a scheduled task. Enter a name such as “Cedar day-end review” and paste the instruction into What to do.
Under When, choose your workdays and a local time when you normally finish. Choose Add, leave it enabled, and use Run now to inspect the first result.
The task runs on the computer. If you close the app or let the computer sleep before that time, do not expect a cloud service to complete the review for you. Local model work can also wait behind another job.
Review the first result against real work
Check one item from each group. For “I can continue,” confirm that the action is still yours. For “waiting,” check who owes the next step. For “status unclear,” decide whether you need another source or a direct question to a colleague.
Look for duplicate tasks described with different wording. Keep one clear entry with the supporting references rather than carrying both into tomorrow's plan.
Do not mark something complete just to make the review shorter. If the source is ambiguous, keep the uncertainty visible and decide the next verification step.
Use the review to prepare tomorrow
Choose the first action yourself after checking the list. Ask for a short starting note that includes the file, conversation, or source you need to open. “Continue the brief” is less useful than “Open the Cedar brief and resolve the delivery-date question in section three.”
Keep waiting items separate so they do not consume the same attention as work you can do immediately. If a follow-up message is needed, request a draft and review its recipient and wording before sending.
The routine's summary is not a promise to monitor other people's progress or automatically chase them. Any such action needs its own supported setup and clear scope.
Fix an unhelpful review
| Symptom | Better next step |
|---|---|
| Too many unrelated items | Restrict project and ownership |
| Completed work appears open | Require the latest completion evidence |
| Waiting items lack a person | Ask for the source or mark owner unknown |
| Routine produces no result | Check enabled state, awake computer, and model queue |
| Important work is absent | Check capture coverage and source access |
Use Edit and Save to improve the instruction after a few checked runs. Keep the source requirement even when you shorten the output.
Schedule one end-of-day review in OGAD and compare its first result with your own list. A checked review gives you a concrete place to start tomorrow without pretending that every loose end is known.

![Why Object.keys(user) Returns string[], Not (keyof User)[]: TypeScript Means It](https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F2q4w48b9bzs43tynyhsf.png)
