One person, one problem, one product

I kept a bullet journal for years, and it was fine for capturing what needed doing but never much good at telling me what to do first. When I heard about time blocking, laying tasks directly against the hours of the day instead of just listing them, I tried it by hand in the same journal, and it fell apart within a couple of weeks. Replanning a page every time something shifted took longer than the list ever had.

Then I found an app that did the time blocking automatically, and for a while it was close to exactly what I needed. Two things ended that. The price crept up past what a single person's calendar was worth, and the company's attention visibly moved toward teams, shipping feature after feature for colleagues coordinating with each other rather than one person planning their own week. So I stopped paying for it and started building a version of my own instead.

That happened a few hours a week, in whatever time was left around other projects, sometimes with whole weeks going by untouched. It became the hardest programming problem I had taken on, and more than once I hit a wall I could not see past and set it aside for months. What changed that was AI-assisted coding, which let me debug and rebuild enough of it that I started relying on it as my actual schedule rather than a side project. My partner started using it next, and the two of us decided together what it still needed. It is at the point now where it might be useful to other people too, so we are opening it up, starting small and honest about it.

Every task gets a score, and the score is built from two numbers you control: how important you said the task is, and how soon it is due. The reason both exist, and the reason the balance between them is a setting rather than a constant, is that people genuinely differ here. Some work to deadlines and want the nearest one first regardless of what it is. Others have important work with no deadline at all, and a deadline-only scheduler buries it forever. Picking one of those models and calling it correct would have been simpler to build and wrong for half the people using it.

Placement is deliberately simple. Tasks are sorted by score and each one goes into the earliest free slot that fits it, with a buffer after it so the day is not packed edge to edge. This is a first-fit approach rather than an optimiser, and that is a real trade: a long task can occupy a large gap that two medium tasks would have used better. A cleverer algorithm would pack the week more tightly, and it would also produce a plan that changes shape for reasons you cannot see. A schedule you can predict is worth more than a schedule that is four percent denser.

Long work is split rather than dropped. A four-hour task will not fit in most real calendars in one piece, so it becomes sittings that fit the gaps you actually have. And when something genuinely cannot be placed, you are told which task it was and why, because a scheduler that hides its failures leaves you trusting a plan that is incomplete.

It is Google Calendar only, so Outlook and iCloud users are not served yet. It is desktop only, because a half-working phone layout would be worse than an honest message telling you to open a laptop. Sync runs one way. It re-plans when you ask it to rather than continuously. The comparison page sets all of that against the alternatives, including the cases where one of them is the better answer for you.

Support, questions and bug reports all reach me directly at info@auto-scheduler.net. Feature requests have their own board once you are signed in, and the full legal details are on the Impressum.