I’ve worked my ass off on SelfManager.ai.
Through all of 2024, all of 2025, and 2026 so far, I’ve put thousands of hours into building it myself, with essentially no marketing budget.
I wanted to build a product that brought together my passion for software development, my experience as an engineer, and my taste in product design. The project has now reached 6,000 commits since I started it in 2016.
I chose to work in software in 2010 because I loved computers. Through freelancing and ABZ Global, I went on to work on hundreds of projects for U.S. clients.
Quality was a requirement. I learned to think past the brief, imagine how people would actually use what I built, and look for ways to give the client a better result than they expected.
Steve Jobs put it in four words. “We don’t ship junk.”CBS News, August 2007
I brought that standard to SelfManager. I also get to live with every decision I make about it. I use it every day, including to plan, track, and review the work of building SelfManager itself.
When an interaction takes an extra step, I’m taking that step too. When I go back to a project weeks later, I’m the person trying to find the decision that explains where I left off. Using the product for years has given me thousands of chances to see where it could work better.
Why I built it around the day
By 2016, after a few years of client work, I kept coming back to time as fundamental to how we experience life. It passes whether we want it to or not. We can learn from the past, think through possible futures, and act today.
My paper system gave each day its own column. Client work, my own projects, and personal commitments all drew on the hours in that column. I wanted to see them together.
That’s what I carried into SelfManager. Each date can hold its own tables. One might be a short list for today. Another might hold part of a client project, along with the notes, files, and conversations you need for the work. Everything Belongs to a Date
The real value of keeping that record became clearer when I looked back across weeks. I could have several busy, productive days in a row and still discover that client work had crowded out development on SelfManager or a personal goal.
The dates showed me how long it had been happening. The time I’d tracked showed where the hours went. The notes I’d written helped me understand why.
Once I could see the pattern, I could change what got my attention today and make a more realistic plan for next week. That idea would later shape how I built AI Review and AI Plan. How I Run a Software Agency and a SaaS in Parallel · A Checked Box Tells You Nothing
What went into the last three years
A lot of the work in 2024 went into things you do with tables every day. Finding one again. Pinning it so it’s easy to open. Moving it to another date. Using it on your phone.
Shared access and comments for the whole table gave people a place to work together. They also changed what other features had to do. Move a table and its conversation should come along. Make a copy and you should be able to choose which parts to bring over.
In early 2025, I counted 126 completed development tasks over the previous seven months. That was the pace of work before some of the biggest releases. SelfManager.ai product history
In the first half of 2025, I started building the period summaries that became AI Review. I added ways to generate tasks from text, summarize a table or a group of pinned tables, and ask follow-up questions about a period of work.
The week and month views also gained more detail about progress and time spent. By the end of the year, I’d added optional table activity logs and quarterly AI reviews. There was now more to work with when you wanted to look back across days or weeks. SelfManager.ai product history
By April 2026, AI Review could use tasks, statuses, completion timestamps, tracked time, and notes. You could choose to include table comments and when they were posted. Where table logging was enabled, the review could also draw on what had been edited and when.
That made questions like “Why does this keep getting pushed back?” more useful. The review had a sequence of events and some of the reasoning behind them.
I introduced AI Plan that April so you could use your recent work as context for what comes next. It drafts editable tables for actual dates ahead. You can go through them and decide what belongs in your plan. April 2026 product update · Introducing AI Plan
April was also when I redesigned the app. I went through the daily workspace, the menus, the dialogs, and the screens you open from them. I worked on how they felt on desktop, how they behaved on a phone, and how they looked in light and dark mode.
I added mobile gestures, rebuilt the notes editor with version history, and gave people more control over the layout. Deadlines, reminders, notifications, and mentions arrived around the same time.
Setting a deadline might take a few seconds. Making it easy to set from a task, easy to find when you’re looking ahead, and able to reach you when it matters took work across several parts of the app. The April redesign · Deadlines, notifications, and mentions
The rebuild you could barely see
In May, I moved SelfManager from Firebase Realtime Database to Cloud Firestore. By then, people already had years of work in the app. A table could hold tasks, notes, comments, AI history, and information about who could access it.
I had to design how those pieces would fit in the new database and work through how to move the data people already had. When someone opened a shared table after the move, they needed to find their work, the conversation, and the right people there.
I was also designing for what I wanted to build next. A task with its own discussion, subtasks, and an assignee is one example. It adds several pieces of information to something that already lives in a table with its own history and permissions.
That’s what made the database rebuild so important to me. I needed a foundation that could handle the features already in use and give me room to build the ones that followed. The May database rebuild
AI agents changed how much of this work I could take on. It felt like I could clone myself ten times.
At any hour I could research a technical question, compare approaches, ask for an explanation, or have an agent challenge a decision I was leaning toward. I used them for coding, reviewing, testing, design iterations, and looking at analytics and user behavior.
That gave me more room to explore how something could work before deciding how it should work. It also gave me more ways to check the result before I shipped it.
What you can do with SelfManager now
From July through September, I added dictation, read-aloud, draft recovery, recurring tables with blueprints, and more detailed team roles. I also added the option to include images in an AI review.
By the October 15 release, tables could hold audio recordings, documents, and video. Documents and transcripts of audio recordings became optional AI context alongside images. I also added a choice of models from Gemini, Claude, GPT, and Grok. The summer v6 releases · The October v7 release
Think about a project where the brief is in a document, the decisions were made on a call, and the follow-ups are in a task list. You can keep those things with the work. When you come back to review it, you can give the AI the brief and the call transcript and ask what was decided and what still needs attention. AI context in the October v7 release
Through MCP, supported assistants can work with the SelfManager tables you give them access to. You can ask an assistant to pull up today’s work or, with permission to make changes, add the next steps from a conversation. SelfManager.ai MCP
The API is live, and webhooks let connected tools respond to events in your tables. I want to keep building out these integrations so the assistants and tools people already use can do more with SelfManager within the access they’ve been given. API and webhooks
Work continued between the big releases. I improved uploads and the video player, made tables move with their files, gave people more control over what a duplicate carries, and added deadlines for whole tables. Those are the kinds of details I keep coming back to because they affect how the app feels in ordinary use. Recent SelfManager.ai updates
What shipped today
Today’s update is live. It upgrades Fable 5.1 to Fable 5.2 and adds three capabilities to tasks.
First, comments directly on a task.
You already have comments for the whole table. With task comments enabled in the table’s settings, feedback and decisions about a specific task can stay with that task. New tables start with the setting off, so you can turn it on where you want it.
Next, subtasks.
Break a task into the steps needed to finish it. A task like “send the proposal” can hold the draft, the pricing check, and the final review. You can open the task and see what’s left.
Finally, task assignments and filtering by teammate.
In a shared table owned by someone on the Team plan, you can assign work and filter by person. See what’s on your plate or look at what a teammate is responsible for.
Invited collaborators can use these features with their existing free or Individual accounts, within their permissions on the table.
If you’re curious about what SelfManager has become, bring whatever you need to plan, organize, or understand. Try it on your own work and see how it fits. Explore SelfManager.ai