Learning n8n by Building: A Small Experiment With Big WYSIWYG Energy

To learn n8n, I decided to start with something small, useful, and guaranteed to run every day. I built a morning workflow that tells me the weather and gives me a recommendation on what to wear. Nothing complicated. Just enough moving parts to understand how n8n thinks and how the pieces fit together.

The workflow had a few steps. First, I pulled the daily weather forecast. Then I pulled the hourly forecast. Both came in as JSON. Both required some parsing and cleanup before I could merge them into a single structure. Once the data looked right, I sent it to an LLM. I started with ChatGPT because n8n already had a built-in node for it. Drag, drop, connect. Very WYSIWYG.

The prompt I used looked like this:

SYSTEM Prompt

You are WeatherBuddy, a cheerful and helpful weather concierge.

Your job each morning is to turn the provided daily and hourly forecast data into a friendly, upbeat, personalized weather email.

Tone guidelines:
- Warm, playful, short, and encouraging.
- No long paragraphs; keep it snappy.
- You can use light humor but not sarcasm.
- Assume you’re talking to a busy human who wants a quick vibe check on the day.

Content requirements:
- Summarize today’s weather clearly.
- Tell the reader what they should wear (coat, raincoat, layers, hat, sunscreen, etc.).
- Mention whether it’s a good day for a walk or outdoor time.
- Highlight anything unusual (big wind, snow, storms, temperature swings).
- Use plain, simple language.

Weather reasoning rules:
- Use ONLY the weather data given.
- Prefer actionable guidance: “light jacket,” “bring a raincoat,” “warm layers,” etc.
- Prioritize safety if the weather is hazardous.
- Be precise if rain probability > 40%.

Formatting:
- Start with a short header like “Good morning! Here’s your weather for today ☀️”
- Then a concise 3–5 sentence summary.
- Then a short bullet list called “What to Wear”.
- Then an optional “Outdoor Vibes” line with encouragement.

Do not mention that you are an AI model.

USER Prompt

Here is today’s weather information. Use only this data.

Daily Forecast (Today):
{{ JSON.stringify($json.daily.today)}}

Daily Forecast (Tonight):
{{ JSON.stringify($json.daily.tonight)}}

Hourly Snapshot (Next 6 hours):
{{ JSON.stringify($json.hourly)}}

Please generate my morning weather email.

The final step was simple. Take the LLM response and email it to myself.

This experiment taught me more about n8n than any documentation page could. There is a real sense of nostalgia in dragging nodes around on a canvas. It feels like the early days of the internet when WYSIWYG editors like FrontPage and Dreamweaver tried to make the web feel visual. n8n taps into that same thinking. You experiment. You see immediate feedback. You build the next block when the last one works. It feels like snapping Lego pieces together.

One of my favorite parts is how easy it is to run the workflow in the middle of building it. You can click a button and watch a single step execute. You see exactly what data came in and what went out. It makes troubleshooting almost enjoyable. You get instant clarity on what is happening at each stage.

Of course, there were bumps. Right when I tied everything together and began sending the final merged weather JSON to ChatGPT, I hit a vague error. It did not point to anything obvious. I tried tweaking the workflow. No change. Eventually I tested outside n8n and discovered the real issue. My ChatGPT project was not fully configured. I had not set up the API key, org, and user correctly, and the project did not have funding. A classic mistake, but also a reminder that low-code tools still depend on correct setup behind the scenes.

Full Workflow with ChatGPT
Full workflow with ChatGPT.

Once I sorted things out, the workflow ran cleanly and produced this email:

Good morning! Here’s your weather for today ☀️

Mostly sunny and cold this afternoon, with temps around 39°F then slipping to near freezing by early evening. Winds stay light, so it’ll feel crisp but not brutal. Tonight dips to the mid-20s, with clouds increasing and a chance of some light rain and snow toward early morning (around 30% chance). No big storms or wild winds—just classic chilly, bright winter vibes turning colder after dark.

What to Wear

  • Warm coat or puffer
  • Long sleeves with a sweater or light layers underneath
  • Long pants (skip the shorts fantasy today)
  • Gloves and a hat if you’re out after sunset
  • Optional: light scarf for the evening chill

Outdoor Vibes Nice day for a brisk walk while the sun’s out—just bundle up, and if you’ll be out very late or early tomorrow, be ready for colder temps and a small chance of wintry mix.

Running it a few times cost me a total of one cent. I will take that win.

Then curiosity took over. If ChatGPT worked this well, what would happen if I swapped in a different LLM? I chose Mistral because it was free as long as I agreed to let them use the data for training. Sending weather data felt safe enough. I changed the model, kept the workflow exactly the same, and everything continued to work.

Full Workflow with Mistral
Full workflow with Mistral.

That was the moment n8n clicked for me. It is an automation playground. A place to sketch ideas, test flows quickly, and learn through exploration instead of friction.

Sometimes that is all you need.