MyTodoBoard: Why I Built My Own Task Board

Most task managers ask me to build a system before I can write down a single task.

Projects, workspaces, boards, statuses, assignees. All of it configured before the first real thing goes in. I have set that up more than once, used it properly for about a week, and then quietly gone back to keeping the important things in my head.

So I built the version I actually wanted. MyTodoBoard is a corkboard. You write a note, give it a priority and a deadline, and pin it. That is the whole model.

Why I built another task manager

There were two reasons, and only one of them was about tasks.

The first is the one above. I did not want a methodology. I wanted a surface where the things that matter are visible and everything else waits its turn.

The second reason is that I have spent eleven years building for other people’s platforms. Plugins inside WordPress, blocks inside Gutenberg, features inside products whose shape somebody else owns. I wanted to carry one thing the whole distance: the front end, the API, the database, the domain, the deploy, and every decision in between.

A task board turned out to be a good candidate. The domain is small enough to hold in your head and awkward enough to stay interesting. Recurring tasks, timezones, overdue rules and sync across devices are all harder than they look from the outside.

A note is three fields

A title, a priority, and a deadline when the thing actually has one.

Everything else is optional and stays out of the way until you want it. A description, tags, steps, attachments, a repeat rule. The dialog does not open with twelve empty fields demanding to be filled before you are allowed to write down “call the printer”.

Priority is the load-bearing idea, because priority is the colour of the paper. Urgent is clay, high is gold, normal is manila, low is sage, someday is lilac. You read the board before you read a word on it.

Two rules do more work than everything else combined:

• Today includes overdue. A task that slipped does not quietly stay in yesterday. It sits in front of you until it is done or deliberately pushed to a new date.
• Completing a note archives it. Nothing is destroyed on the way past. The Archive is where I go when I need to convince myself the week was not wasted.

Three ways to look at the same board

The same notes render three ways, and the choice is one setting.

The sticky board is the corkboard. The modern board is the same data as five priority lanes on a blueprint grid, where the layout is the status report. The calendar is a week grid where a task sits on its day and a timed one sits at its hour.

The calendar forced the decision I am happiest with. A booked block of time looked like a new kind of object, and modelling it as one would have meant a new table, new endpoints, and a second implementation of search, tags, recurrence, undo and export.

Instead a schedule is a note that happens to carry a start time and an end time. Every surface that already understood notes understood schedules on the first day, at a cost of three fields. A weekly ten o’clock recurs correctly because the code that spawns the next occurrence never had to learn that anything new existed.

That is the kind of decision that does not show up in a screenshot and decides how much the next feature costs.

Letting an assistant use the board

The part I did not expect to enjoy most is that the board has an address an AI assistant can talk to.

MyTodoBoard speaks the Model Context Protocol. Claude and ChatGPT take the address as a connector and ask you to sign in. Claude Code, Cursor and scripts cannot show you a sign-in page, so they use a key you create yourself, scoped to what you are willing to let them do: read the board, change the board, see your profile.

The key is shown once, it is named so you can tell your laptop from your work machine, and it can be revoked without touching anything else.

One rule matters more than the tooling. An assistant filling in a board has to guess constantly. “Invoice by Friday” means a date, and “urgent” in a sentence may or may not mean the urgent lane. So the contract is that it declares what it inferred and quotes the words it read it from, or it asks. A board quietly filled with confident guesses is worse than an empty one, because you stop trusting the colours.

Where it fits, and where it does not

It suits a narrow set of situations well:

• one person’s working day, not a team’s sprint
• work that arrives faster than it can be organised
• a week you need to see before you agree to anything else
• talking a day’s tasks to an assistant instead of typing them in

And it is honest about the rest. There are no teams, no assignees, no projects and no billing. There is no offline mode yet, so it installs to a phone but will not help you on a plane. Attachments still travel inside the note rather than sitting in object storage, which is fine for a scanned invoice and wrong for anything larger.

Without an account the board lives in the browser you opened it in. That is a real mode rather than a trial. Signing in moves it to an account, syncs it across devices, and is what makes the assistant integration possible at all.

Takeaways

Building the whole stack myself changed what I paid attention to. The features that took longest were never the visible ones. Recurrence, timezone handling, what happens to a note when the day rolls over, and what an assistant is allowed to assume were all more expensive than any view in this post.

The corkboard was the easy part. The rules underneath it were the product.

If you are building something small for yourself, I would pick a domain you already have opinions about. I have used task boards for years, so I did not have to guess what was worth arguing over. That saved more time than any framework choice.