The next step, then, is to create a command node in Tana.
You can create a command node by placing your cursor in a node, hitting Cmd/Ctrl+k and searching for "Convert to command node".
The text you put into the node is going to be name of the command:
Once you've converted the node to a command node, you can add actions (Tana calls them commands) that will happen when you run the command (node) as well as "parameters" i.e. settings.
A command node can perform multiple actions/commands in sequence!
Add actions by typing "@":
We're going to add the "Ask AI" command to our command node and then click the "Show configuration" button.
This will allow us to set a couple of parameters (settings) for our Ask AI command by typing > and adding some fields.
The first one we're going to add is the prompt:
The prompt is obviously the most important thing – it's in the template I'll share at the end of the thread :)
This is one that works for me right now – but I'm always tinkering with it, and so should you.
Basically what I'm asking GPT to do here is to write me 1) a summary of everything that I've said and 2) go through and get the a) main points of my monolog, b) action items I'm mentioning, c) ask me follow-up questions and d) give me some pushback on what I've said.
To test your prompt, you want to make heavy use of Tana's great "Prompt Workbench".
It's a convenient way to select one node from your graph and test your prompt on it.
You can also set a parameter called "Temperature (OpenAI)" on your AskAI command.
This is roughly how "creative" GPT gets with its execution. Right now I've set mine really low (0.1) – tune this to your liking.
And with that, you're ready to run your new command and have it summarize your voice notes!
One thing to note here is that this command doesn't transcribe the note for you – I use Tana's "transcribe" button for that.
Run this command once you've transcribed a note and get this:
And if you want a powerful and fully done-for-you Second Brain in Tana, check out my template Tanarian Brain! cortexfutura.com/tanarian-brain
Enjoyed this thread?
If you use Tana, you'll love my template Tanarian Brain! It's a fully featured implementation of a Second Brain in Tana and comes with built-in support for PARA, Zettelkasten, GTD, AND has AI built into it too!
Check it out here: cortexfutura.com/tanarian-brain
• • •
Missing some Tweet in this thread? You can try to
force a refresh
Avoid High Modernist Design in your Knowledge Base
Whether you're using @tana_inc, Notion, or any other tool to manage knowledge and collaborate – that's the one thing you want to avoid.
Let me tell you what I mean: 🧵
What is "High Modernism"?
It's a term from James Scott's fantastic book "Seeing Like A State" and describes, slightly condensing,
"uncritical optimism about comprehensive planning."
It's been applied to everything from forestry to city planning – and you should avoid it.
So what does it have to do with team knowledge bases?
When we decide to create a system to document things, to build a wiki, to collaborate, it's very easy to start by _designing_ such a system from scratch.
1) Work in 2h blocks 2) Use noise cancelling headphones 3) Use 4) Interstitial Journaling 5) Rapid Feedback
Some more detail:
1) Work in 2h blocks
I did this way before Huberman taught me about ultradian cycles, but apparently our brain doesn't just cycle through 1.5h long periods while we sleep, but also while we are awake.
So what I do is the following (starting at top of hour for ease of demonstration):
08:00 – Plan: what am I going to focus on, what's the goal? How am I doing, any interruptions I need to expect? How will I handle them?
08:10 – Work for 30min
08:40 – 5min break: stand up, look around, stretch, refocus: where do I move my focus next?
08:45 – Work for 30min
09:15 – 5min break: stand up, look around, stretch, refocus: where do I move my focus next?
09:20 – Work for 30min
09:50 – Review: how did this round go? What did I get done? Anything I can do better? Move, take a short break.
Ready to go again at 10:00
I've done full 8-12h days like this, but often even a single 2h cycle gets me deep enough into the zone that I stop following the template until I physically can't anymore or am interrupted.
Also works at the end of the day to get that one more rep in.
2) Noise cancelling headphones are amazing for focus even without any music in them.
3) is magic I don't understand but use liberally. Best with noise cancelling headphones
4) Interstitial Journalling
This is what I do during my 2h sessions and esp. the breaks: plan/review/refocus in writing so that you a) can think better and b) have a record you can go through later. Flows directly into
5) Using this approach gives you a rapid feedback cycle in 30min and 2h chunks. If you do this for a week, you're going to learn A LOT about how you work, what the most common interruptions and friction points are, and how you can eliminate them.
Additional notes:
This obviously works best with a maker schedule and less well with a manager schedule. If your whole day is meetings and you don't have interruption-free 2h blocks that's not great.
This is not something to slavishly adhere to for the sake of following the template/program. I start a coding/reading/writing session like this, but I'm not forcing myself to stop if I'm deep in the zone.
This is great for learning scoping: you usually vastly overestimate what you can get done in 30min. After doing this for a while you'll get a very intuitive sense for how long certain tasks take you.brain.fm
Back in the day when I worked for Ultraworking we used to teach this as "Work Cycles" and the results people got from it were really, really good.
We also used to host "Work Marathons", a three day event where you'd try to pack as much focused work into the 72h as possible (without compromising on sleep). People regularly worked three 12h days and got a LOT of work done.
Now Ultraworking doesn't exist anymore as far as I know, but their website and template for doing "Work Cycles" are still up:
I've also done a video on how I've set this up in Tana from a year ago (man, how time flies):
"No two kitchens are the same, therefore you need to build your kitchen cabinets from scratch yourself" - @fortelabs on Second Brain templates (paraphrased)
That's basically what Tiago wrote in a newsletter he sent out yesterday
Now of course I'm biased, but here's what I think
OF COURSE you need to modify any template to fit it exactly to your needs. Even a big, encompassing template like Tanarian Brain isn't going to work 100% perfect out of the box for you.
And everyone who's bought it knows and expects that.
That's why Tanarian Brain comes with
A) loads of videos to show you how it works so you can modify it and
B) an active support community where people discuss their modifications and you can ask for help
The biggest unacknowledged problem for proponents of the Zettelkasten method as a tool for writing is that its inventor, Niklas Luhmann, is famous for how badly he wrote.
I still believe a Zettelkasten is fantastic tool, but only if you learn how to go from non-linear to linear.
This is a much better linearization of Luhmann's ideas (well, some of them, at least) than Luhmann has ever produced himself:
This is also, I believe, a reason why shareable knowledge graphs (so a published Roam Graph or a published Tana workspace) are ONLY useful when they contain some form of linearization of the material presented in the graph.
In Tana you can link from any node to any other node in your workspace very easily. You type @ and then search for the node you want to link to.
But sometimes "transcluding" the full text isn't what you want!
That's where alias links come in.
Sometimes you don't want to have the whole text there, however: our example reads very weirdly at the moment.
What we want instead is something like "As Hamlet has remarked" and have the "has remarked" link to the "To be" node.
Here's how the finished product will look like:
To do this, we use what Tana calls an "alias" – it keeps the link, but changes the text that we see.
Select the "transclusion" (i.e. the "To be, or not to be,..." behind "As Hamlet") and then type Cmd/Ctrl+k. Then search for "alias" and select "Set alias to".