For over a year, every task I add to a list on Google Tasks has a best-describing emoji assigned as a prefix to its title. I liked the methodology so much I even automated this process to make sure all my tasks have emojis. Why is that, and does it change anything in my day-to-day life?

Readability

Well-chosen icons near frequent action items can improve the UX, drastically. At least for me, I won’t dive into scientific research nor professional UI guidelines. The choice isn’t so obvious, as you can see: Apple’s macOS includes icons in context menus, while Firefox and GNOME don’t.

Getting lost in plain text

I keep track of a few task lists; with many entries, it all becomes cluttered with plain text, hard to navigate. Normally, I don’t have a need to search for tasks, but even if I needed to, Google Tasks does not provide such a feature!

The easiest way, then, is just to scan for tasks in a given list. Here’s a little example of how such a list looks before emojification and after.

It’s a small sample, but it scales. When you have dozens of tasks, it quickly becomes visually hard to see through.

I like how I can see the idea of a task and what it’s related to just by looking at it.

Assigning emojis by hand

At the beginning, I assigned every emoji before a new task’s title by hand. Generally, it was quite simple and fast - think of an emoji, pick it, save it.

The problem was… the time and attention needed to find the one icon and put it there. You just want to write down a thing and hide your phone back to the pocket. The outcome was contradictory. I saved milliseconds on reading by spending seconds on writing.

Then I remembered the ongoing automation trend, especially with n8n. I almost missed it all because I simply didn’t have anything more to automate as I think my workflows are pretty simple and do not need all this extra processing.

Automation

A simple workflow gathers all the tasks from my Google account and assigns the best-fitting emojis as prefixes for their titles. What would it be without the use of AI?

The entire job runs every 5 minutes and looks like this:

1. Get all task lists and their tasks
2. Assign the parent list identifier to every task
3. Filter tasks which already have an emoji at the beginning of the title
4. Ask an AI model for the best-matching emoji and prefix the name
5. Split the resulting structured JSON and update every task

Keeping the rent low

You may ask why it’s done this way and not another. There’s adding a task list identifier to a task, joining, then splitting… There are two main factors that lead up to this:

1. Rate limits - to keep the costs minimal, I prefer to ask the model once for an answer and perform a bulk update. A single request has all the needed data. No need - no request.
2. Support for multiple task lists - Making a single request requires including an identifier of a list the task is coming from.

This way I can easily run the workflow multiple times an hour with a total cost of $0. Without hitting any limits on the Google Tasks API or the Gemini API side.

A glimpse on the JSON structure

Just to demonstrate how the mentioned single request looks in practice.

"input": [
  {
    "title": "An example",
    "taskId": "aUxrSHFEVlRIYm8xT0hUOQ",
    "tasklistId": "MzQ2MjYwMzY2ODM1NDIwMzc0MTM6MDow"
  }
]

"output": [
  {
    "title": "📝 An example",
    "taskId": "aUxrSHFEVlRIYm8xT0hUOQ",
    "tasklistId": "MzQ2MjYwMzY2ODM1NDIwMzc0MTM6MDow"
  }
]

Yes, this solution is wasting tokens on processing the identifiers, etc., but at this scale it does not matter at all. The per-request limits are more painful than tokens.

And yes, you’re giving your task names to a big and evil corporation to be processed by their artificial intelligence. At least it’s still under the same umbrella as the task manager and calendar.

It’s not ideal, technically

The Google Tasks authorization needs to be refreshed every week as OAuth apps without verification can’t work any longer. Now I don’t mind, but the need to log in to the n8n credentials manager and refresh the connection is real.

There were also times when some updates to n8n itself were breaking the workflow. Fortunately, the fixes were pretty simple; nonetheless, it shows how fragile the solution actually is.

What if not AI

More as a sidenote, but I was thinking about how to pick the best emoji for my task without involving LLMs as it feels like shooting at a fly with a bazooka.

Unfortunately, it wasn’t that simple. You could build some sort of dictionary of keywords and emoji names, and match them with titles. For sure there are ways to make it work. This uncertainty led to the use of Gemini. A light model is ideal for a job like this - classify action, find corresponding icon, assign.

The other way I’d think of, without LLMs, is to use something like EmojiDB, but as far as I know they do not provide a public API. To make it work you’d need to write a scraper and query the service with your task name or its elements.

Then you remember there are more languages in the world than English.

Icons are nice, not only in task titles

I started this post by noticing action items, context menus and general UX. This post is heavily focused on emojis and tasks; I feel it. Seizing the opportunity, I’d like to give you a few examples where icons / emojis are nice too and where I use them too.

• Obsidian - with the use of the Iconize plugin, I assign icons to my folders and important notes. I like the look of them and they work for me the same way emojis in task names do, instantly seeing what it’s about.
• Calendar - this one is purely aesthetic; most of the events in my calendar have emojis prefixed to the title as I like the look. Particularly with At a Glance. Not a rule though.
• Attention grabbers - mostly in text, when I want to grab readers’ attention and give a specific feeling. An example of this would be prepending an important message with “⚠️”. Personally, I do this only at work.

Not everything needs an emoji

It’s not like I add an emoticon to every checkbox I write. I do not include emojis in the GitHub issues or Jira tickets I report. I do this where I find it viable to provide a value, whether it’s an aesthetic or practical one.

Back in the day, I loved adding emojis to headers in repositories’ README files. I found this eye-pleasing and good for navigation when you could see a little “🏗️” icon next to the “Building” section of a document.

It has changed, and now I prefer clear and concise descriptions. Notably in documentation, and I feel like such a README is one. I can also blame the masses of AI-slop produced every hour and their distinctive way of writing, often full of emojis. Emojis are very easy to overdose on and produce visual clutter.

It’s all personal

This post touched on at least a few different topics. It was(n’t) intended. A few things needed to be pointed out, and about some I just forgot. The key takeaway should be:

This is not a tutorial nor a recommendation. I wanted to show what, why and how I do the emojis. Maybe you will find something helpful here and try it yourself!

See ya!