I attended my third vibe coding meetup today. I could not attend the last one because I was not sure I had anything to build. Yes, I need to work on that.

The event today, was graciously hosted by Antilooppi at the pool lounge. This is next to McDonald’s at Hakaniemi, my old place. As I walked through the area to get to the venue, it jogged some old memories for me. The venue was beautiful, with beautiful furniture and classy interiors. There was a food station in one corner and wide windows on the other wall.

Jesse and the Vibe Coding Finland team seem to be outdoing themselves each time.

There were three speakers today and all three of them left me with things to think and read further about. Yes, it was that kind of an evening!

1. Vibe Meets Design

Kimmo Ollikainen was up first, with his talk on using AI for design tasks. He talked about the double diamond design process, which is basically:

Research > analyze > design > deliver

We might be building the wrong thing if we just start work and iterate. So need to be clear about what we are building.

Where to use AI?

1. Research phase
2. Automate people with AI
3. Design with AI
   1. Might be better to use lower fidelity - signals that changes can be made and that it’s not already done
4. Coding with AI
   1. Rapid experimentation as part of design

This was a designer’s presentation - beautiful, matching fonts, fun language. I enjoyed it very much.

2. Real Back office Al @Retta by Zeal Al

Next up was Mikko Valtonen, talking about implementing an AI system to help with intelligent invoicing. He left me with an idea about forward deployment engineers.

1. Future is you don’t write the code, you write the prompts - software 3.0 by Andrej.
2. You need to build a good ui for verification of the model’s output.
3. Forward deployment engineer - sit with the end user, learn root causes, do design, build demos, get feedback, improve.

3. A recovering vibe coder

Finally, Janne Ruohisto took the stage and told a personal story about using AI to build a LMS.

He left me with a wonderful picture of AI being a bull in a china shop - can to destroy stuff while building/fixing something elsewhere.

And also with an idea of a continuous learning process with AI, where after every mess up, you discuss with the agent what went wrong, continuous learning type thing.