No-Code 101: Node-based thinking

Playing software LEGO

This post is part of the No-Code 101 series, a quick start, no nonsense guide to nontechnical people to learn how to build automations and AI agents without learning to code.

---

Most no-code tools use some kind of drag and drop node-based interface. This means that each automation:

• involves a series of steps following some logic

• each step is a self-contained unit of input, task, output

• third party apps (like Gmail, OpenAI or Instagram) exist as a variety of these steps you can choose from.

This becomes clear when you put a few of these tools together to see how workflows look like. These tools are basically flowchart generators, with the only extra that they can also execute the flowcharts.

This node-based thinking, where these interconnected nodes process data in some smart way is also called dataflow programming.

Always try to think of these building blocks. It takes a while to get used to it, but you can use my Node-based Thinking Tutor to help you nail it:

Click here for the Node-based Thinking AI Tutor. (FREE)

You should know

There are a couple of ideas in Dataflow that you should be aware of:

• You’re sending packets of data from one node to another in a predefined order. This is why we call them data packets.

• Your flow can execute via push or pull flows:

  • Push Execution means the flow (or the node) fires when the input becomes available. For example in Make, these will be modules that have a little “Instant” tag next to them. This is mostly how Zapier’s Zaps run.¹

  • Pull Execution means the flow (or the node) fires when output is requested. This is mostly how Make scenarios run based schedule or webhook triggers.

• Execution can by synchronous (sync) or asynchronous (async).

  • Sync execution means that completing each node will take a unit of time and each node will wait for the others to finish working, so the fast nodes slow down to let slow ones catch up.

  • Async execution means that nodes never wait for others, so everyone runs at full speed.

Which no-code tool should I use?

Selecting your no-code tool is basically a matter of preference and budget. Their capabilities are similar and converging. Zapier just announced Canvas, Make is working on Grid. Both Zapier and Make have AI-assisted Zap/scenario generation and n8n is reportedly be working on it too.

Here’s my history of using these tools:

1. First I started with IFTTT in 2014. I remember the very first automation I built: auto-archive every photo I took with my iPhone to a Dropbox folder. If learning these proves to be a challenge for you, start with IFTTT. It’s designed for the simplest automations but will give you a good feel for this.

2. Then I used Zapier in most of my businesses, including my first AI startup, Brandvee until 2017.

3. At Promptmaster I became a power user of Make, partnered up with them and even did some training to become an official Make trainer. (I didn’t, left the company before I could finish it).

4. During my last months at my old company I was already toying with n8n but only made the full switch while developing High Signal.

As you can see I’ve been using no-code tools for over a decade. I only switched when I needed extra functionality.

Make handles diverging paths (forks) in your workflow more gracefully than Zapier, so when I started building more scenarios that required forks, I switched.

However, Make doesn’t understand loops and has a wonky way of handling data transformations. I can use ChatGPT to write JavaScript expressions for me to transform let’s say one data format to another, but Make’s custom functions is a premium feature. I started to paying top dollar for Make and it wasn’t worth it anymore.

n8n is now self-hosted on my own server for $5/month and that’s all the cost I need to pay for it. It also supports JS expressions natively and handles loops gracefully.

So I suggest you use n8n.

1. This is not entirely correct, I simplified things. Technically, Zapier polls the original data source every two minutes, meaning it runs a secondary pull execution requesting for new input to be passed along. If there is new input, then that triggers your Zap.