A Checked Box Tells You Nothing - How Marian Sorca Uses One Table to Capture Everything and Let AI Review It Every Week

Most task managers give you a graveyard of checkboxes at the end of the week.

Write proposal.
Client call.
Fix homepage bug.

Great. But why did the proposal take three days instead of one? What was actually decided on that call? What did you try before the fix that did not work?

A checked box has no memory. And that means your AI weekly review is just counting completed tasks, which is useless for actual reflection.

I've been using a different system for the past 9 months. This article explains exactly how it works.

My name is Marian Sorca and I'm the founder of SelfManager.ai. What I'm sharing here is my actual daily workflow, not a feature list.

The problem with keeping tasks and context in separate apps

Most people have some version of this setup:

• tasks in a task manager
• notes or journaling in a separate app
• screenshots scattered across folders
• music and media queued up somewhere else
• reviews happening manually, once a month, if at all

The friction between all those places is where context dies. By the time you sit down for a weekly review, you're working from memory, and memory is unreliable.

The fix is not a better note-taking app or a more powerful task manager. The fix is having everything for a day live in one place, so that when you review, the data is already there waiting for you.

The core idea: one table, one day

In SelfManager.ai, each table belongs to a calendar day. You can have multiple tables per day, one per project, one for personal tasks, one for anything else, but the day is always the container.

That one structural decision changes how data accumulates over time. By the end of a week, you do not have a list of tasks. You have a layered record of what actually happened.

Here's what a single table holds for me.

1. Tasks with time tracking and linked tables

Each task has its own timer. I track time at the task level, not the project level, which means my weekly AI review can tell me not just what I completed but how long each thing actually took versus how long I expected.

Inside any task, I can also paste a link to another table. Clicking it opens that table instantly with full context. If a piece of work started on Monday but continued on Thursday, those two days are connected, one click apart.

2. Duplicate and transfer to the next day

If I have a recurring project structure or a setup I want to reuse, I duplicate the table and move it to the next relevant day in seconds. No rebuilding, no reformatting. The structure carries over and I just update what has changed.

This is especially useful for ongoing client work where the task structure stays the same but the specific items change week to week.

3. Comments as a living journal and media player

This is the feature that surprises people most.

The comments section of a table is where I put everything that is not a task but still belongs to that day. Specifically:

• music I want to listen to while working, YouTube links that play directly inside the app without ever opening a new tab
• interviews and podcasts I've queued up, same thing, they play inside SelfManager.ai
• journaling, what stood out that day, why I made a specific decision, what a conversation actually led to, what I'd do differently

That last point is the one that matters most for reviews. When I write "decided to pause the campaign because landing page conversion was below 2% - will revisit after A/B test results" in a comment, that reasoning lives right next to the tasks it belongs to.

Not in a separate notes page. Not in a journal app I'll open twice a year. Right there, attached to the day and the work.

When AI reviews my week at the end of Sunday, it reads those comments. It knows why something got delayed. It knows what I was thinking when I made a call that turned out to be wrong. It can give me a qualitative summary of the week, not just a count of what I checked off.

4. Images at full resolution, effectively unlimited

One table stores up to 100 images without any compression. If I need more, I create another table. There is also a dedicated page in the app that shows every image I've ever uploaded, organized chronologically by date.

In practice I use this for:

• screenshots of results, graphs, or data I want to remember
• design drafts at a specific stage
• receipts or references attached to the day they were relevant
• anything visual that gives future-me context about what was happening

One screenshot can bring back the entire context of a decision better than three paragraphs of notes. Combined with the comments journaling, this is what makes monthly and quarterly reviews actually meaningful instead of just abstract.

5. Notes per table

Separate from comments, each table also has a notes field. I use this for structured reference information about a project, things I want to be able to scan quickly rather than scroll through a comment thread to find.

Comments are for chronological context. Notes are for stable reference.

Then, at the end of every Sunday

I open the AI Period Summary, select the week, and let it read all of it, tasks, time tracking, comments, journal entries, notes, everything.

What comes back is not "you completed 34 tasks this week." It is a qualitative picture of how the week actually went. The AI has context because I gave it context throughout the week, one small note at a time.

I follow up with questions like:

• What decisions did I make this week and do they still make sense?
• What kept getting postponed and what was the reason each time?
• What patterns do you notice across this week?
• Turn this into a priority plan for next week.

I do the same at the end of every month and quarter. The quarterly reviews in particular have become something I look forward to, because after 9 months of logging this way, the data is rich enough to show real patterns, not just recent noise.

Why this works when other systems do not

Most productivity systems fail at the review stage because they only captured the output, the checkboxes, and not the process. So when you sit down to review, you're reconstructing from memory what the week was like. That reconstruction is almost always incomplete and often inaccurate.

When your context lives next to your tasks, in the same place, on the same day, the review writes itself. The AI is not guessing. It is reading what you actually wrote, in the moment you wrote it, about work that was fresh in your mind.

That is the difference between a review that tells you something useful and one that just makes you feel briefly organized before the next week swallows everything again.

The goal was never more checkboxes. The goal was better weeks, and you can only build those if you understand the one that just ended.

SelfManager.ai is free to try for 7 days, no payment info required. Individual plan is $8/month.