Yesterday, there was some discussion around how to keep track of to-do's across multiple teams/meetings/projects without running into Supertag overload in @tana_inc.

In order to structure this, I use the same context field on both my meetings and to-do supertag.

This context field get automatically filled if there is an ancestor with the same field. So if I create a to-do in a meeting with the context "Team A", the context of the to-do is automatically set to this.


For this to work, you will need to create a context Supertag and ensure that everything that can be a context (project/team/meeting) extends that tag.

This will leave you with the feature as shown in the animation below, where contexts are filled without me having to do anything.

You can even upgrade the Context field to be an AI enhanced field to auto get the context from the title of the meeting, but I just set it myself.

Organizing To-Do's Across Multiple Teams

I have some tasks that always have the same sub-tasks in @tana_inc. An example of this is when someone calls in sick, I need to remember to also mark them as recovered with HR. Another example could be a multi-step salary approval process.

This ensures, I never lose track.

In the example, you can see I used title expressions on the "salary-request" Supertag to specify how it should look like.

I also added the three default sub-tasks in the template for the Supertag.

Now, the hidden trick is that every node secretly supports title expressions. if we CMD+K → "Debug node" over one of the sub-tasks in the template, we can find a "Title expression" field.

Task Templates for Consistent Sub-Tasks

During work, we often feel either too busy or not busy enough. Given the dotted line is our maximum workload, we oscillate above and below it.

This is because our workload usually cannot be adjusted on the spot. It takes time and when we aren't that busy, we take on too much.


We could take on much less and ensure the peaks never go above our maximum workload, but that would mean that during the valleys we really do too little. So the best we can strive for is to reduce the amplitude over time.


I always sort of knew this, but I was listening to Deep Questions with Cal Newport and he verbalised this so well. He also suggested steps to take when you are overloaded.

When you are in a peak, he suggests to do three things, **before** you get to work:

1. Reduce what you can right away; Move, Cancel, Simplify
2. Make a plan for the work; This doesn't reduce the work, but makes it feel manageable.
3. Plan to prevent the same in the future

Number 3 feels counter intuitive, but it gives a huge psychological boost and in the end is the only thing that will reduce the amplitude of your peaks.

Listen to the full podcast on confronting overload here: https://share.snipd.com/episode/963424a9-f4e3-41cf-8aed-b161a4cb8392

Balancing Workload Amplitude

Yesterday, we started planning our January trip. We will add another field to the table, the airport code, convert it to an AI field, and remove the quotes.

Now let's see about flight prices. Using @serp_api, you get a 100 free API searches, and they include Google Flights!

We can create a command to use the @serp_api in @tana_inc to look up flight prices. First, add a node "Get flight prices", convert it to a command node, and add a "Make API call" command.

Add a flight price field to the table and set it as the target.

Note that I added the the names of the fields in the URL as parameters using the same format as we use in title expressions.

I also used a trick, when you add ${secret:NAME}, it will store the secret in your settings, so you can reuse it later on.

I put the date and departure date hardcoded in the URL for now. And when we run the command on all of the nodes in our proposed cities.

Most worked, one didn't, but we can ignore that for now. Now, create another quick AI command to extract the number and we are done.

This is much faster than going into Google Flights myself and checking all the prices, plus when I change something, I can update it all.

Naturally, you can build upon this and make it much more fancy if you want :)

Planning a January Trip: Flight Prices and API Integration

One of the way I use @tana_inc is for travel planning. This January we plan to go someplace to work. First, let's ask the AI for some suggestions with CMD+K → "Ask AI"


One of the common issues is that the AI tends to number them. I don't want that, so I select all nodes and CMD+K → "Ask AI to", then I type "Remove all numbers". Then I hit backspace to remove the original selected nodes. We have a clean list now.


I now CMD+K, "View as table", and change the title to "January Travel Plans", and add a "Country" field. Then I CMD+K over the field to upgrade it to an "AI lookup field". And run it.

Ugh. This happens a lot. The AI adds quotes around the strings. Let's fix this.

Add a node above named "Remove quotes", then CMD+K on the node and "Convert it to an AI command", edit it and add the field "Insert output strategy" and select "Replace contents".

Select all the countries and run the newly created AI command to fix the issue.

I normally create a command and just remove it again, but this one I kept as it occurs quite a lot. We will continue tomorrow :)

Travel Planning with @tana_inc

Not all of the ✨ AI commands require a lot of setup. There is an easy to use shortcut CMD+K action that most people don't know about in @tana_inc. It allows you to quickly create a custom AI command, with minimal effort.


Write your prompt in a new node, CMD+K → "Convert to AI command", and done. It even uses AI to generate a nice command name for you.


This is a great starting point to learn more, as you can then proceed to edit the command and tweak the prompt to get it to do what you want.

I'd start by changing the model to GPT-4-Turbo, by adding the "Model" field (type "> model" to find it)

Easily Create AI Commands with CMD+K Shortcut

The @tana_inc "Ask AI" command we found earlier, can also be used in our own commands. Let's create an example of this to scratch one of my itches: summarising a webpage.

We start again, by creating a new node. Title it "Summarise URL", and CMD+K → "Convert to command node"

Then add "Make API request", this is a very powerful one, but for now we are just using it to get the contents of a webpage. We also set the "Parse result" to convert the HTML to Tana paste, and we set the "Target node".

We set the URL to ${name}: the value of the current node.


We do this last step to ensure the contents of the page are cleanly placed into @tana_inc, and because we want to remove them later (we are only interested in the summary).

Next, we add the "Ask AI" command to summarise the "Content" field we just filled, and set its model.

Finally, we remove the "Content" field again. And now we can CMD+K on any URL to summarise its contents underneath it!

Commands are immensely powerful building blocks to create advanced features in @tana_inc yourself.

Check out one of my older videos on commands: https://youtu.be/8KddAyC5hAk?si=rBDpyJRpwTVq6VPr

Creating a Webpage Summarizer Command with @tana_inc "Ask AI"

Now that we have a basic understanding of how commands work, let's  involve some AI ✨  again.

In order to play with this, we return to our #podcast tag, and let's add a transcript.

Theoretically, you could download the audio file, upload it to @tana_inc, and transcribe it, but for large bodies of text, I usually prefer to find a transcript online.

As I am a member of @farnamstreet, I can download the transcript from the last podcast I listened to there.

First, we add the "Transcript" field to the #podcast supertag, and set it to hide when empty. Then we paste our transcript into @tana_inc.

Especially for fields with lots of content, I often configure their "Page size". CMD+K "Configure Field" on the transcript field, then under "Advanced", set "Page size" to a number (of nodes).

This makes it all more legible.

Then we add another Field, titled "Suggestions", and make it an AI enhanced field, and fill the prompt with what we want suggestions on.

Under the hood, this actually uses a command too, you can see (and edit) it under advanced. As you can see, I updated the model here.

Playing with AI and Podcast Transcripts

In @tana_inc, you can automate certain actions with commands. In my last thread, I showed how Mail to Tana extracted to-do suggestions from my emails. Let's create two commands to either accept or reject the suggestions.

Start by creating a new node titled "Accept", and then CMD+K, "Convert to command". The icon in front will change. Next let's see what we want our command to do.

To accept a suggestion, we want to tag it with #todo and remove the #suggestion tag.

Under the newly created command node, @ search for "Add tags", and then "Remove tags".

Then, click "Add configuration", and under add @ search for the #todo Supertag, and under remove @ search for the #suggestion tag.

Now, when you CMD+K over the suggestion, you can search for "Accept", and the automated actions will happen. The reject command is even simpler: just remove the suggestion tag.

Next, we configure the #suggestion Supertag, scroll all the way down to "Advanced", and @ search for Accept and Reject under "Commands".

This will add two buttons next to every suggested to-do, so you can easily trigger each command.

This show the enormous hidden power in @tana_inc as you start to explore automations through commands.

I suggest you move your commands to a central spot to keep them all organised.

Automating Actions with Commands

When I listen to podcasts or books, I always struggled to capture their learnings. I have noticed I need to be active listening, preferably with a laptop and @tana_inc next to me. This takes a lot of time, but it is worth it.

I listen to my podcasts in the car. Using @snipd_app I mark the sections that move me. At a later point, I will manually go over each section and capture them in @tana_inc.

I tried @readwise, but the output is just not good enough, plus going over the sections is helpful.

I start by creating a node and tagging it with "podcast" and filling in the fields I want to keep track of. These are mostly to satisfy my need for structure.

I use a title expression in #podcast to ensure each episode looks the same.

In the @snipd_app, I will go through each of my 'snips', re-listen to them, and extract their key learnings. The AI generated notes rarely match what I heard in the sections, not because the transcripts are wrong (far from it), but because they trigger different things in me.


I love using the inline feature of @tana_inc here, SHIFT-Click any reference to open it in a box without leaving your context. This way you can edit in multiple locations at the same time.


Finally, I add my take-aways as contextual nodes on the day page (using CMD+OPTION+↓)