This week’s update will be brief. TLDR: I’m on track for the August MVP deadline and actually managed to get more done than I’d planned. However, just as I was beginning to feel confident, I got a timely reminder that I'm still heavily dependent on LLMs when building software.

Pando is starting to feel real

This week, I completed the core MVP workflow for Pando and built the first version of the UI. For the first time, you can move through the app, create opportunities and navigate between the main screens. I wasn't expecting to get this far, so it felt great to hit this milestone.

This is the first time the app feels like the product I'm working towards putting in front of users.

Progress on learning to code

Switching my coding sessions to first thing in the morning has been a great decision, and I'm continuing to make steady progress each day.

I’m on track to finish the JavaScript course before the end of July, after which I’ll begin learning TypeScript.

The biggest change I've noticed is how I’m now working with AI. I'm increasingly able to read the code, contribute to implementation decisions and make small changes myself rather than simply accepting whatever the model generates. Compared with my first Next.js project just a couple of months ago, that represents significant progress.

This week's biggest lesson

Just as the week was finishing, I got a serious reminder of how much I still have to learn.

I was implementing TanStack Table when I introduced a bug that caused every link on the page to stop working after deleting a row. At first I assumed it would be a quick fix, but I quickly realised I didn't even know where to begin debugging. There wasn't an obvious error message to investigate or a clear prompt I could give an LLM to get to the root cause. It took me a couple of hours of going round in circles before I eventually isolated the issue and fixed it.

It was a timely reminder that I'm still heavily dependent on AI to build. When something falls outside what the model can easily deal with, I don't yet have the experience to know where to investigate.

I got this issue solved so there was no harm done, but it reinforced why I'm investing so much time learning to code myself. Once you move beyond the happy path, being able to reason about the code becomes far more valuable than simply being able to generate it.

You don't know what you don't know. As a solo founder, I want to uncover as many of those unknowns as possible before they're discovered by my users instead of me.