AI Tools: Beyond The Hype

These claims around AI coding tools are getting wild. I want to see for myself what’s behind the hype.

Having switched between coding and management roles multiple times, each transition forced me to relearn how to build efficiently.

I’ve developed an interest in finding the right tools and workflows that deliver results.

Here’s my current flow.

I start sketching on a whiteboard.
Then refine in Figma.
Then start coding in Neovim (I switched from VSCode, which I also liked - that’s another story).

When coding, I have two main flows:

1. Claude 3.5 Sonnet + Aider
   Claude maps my entire git repo and edits multiple files in one go. I can stay in my editor.

2. Manual back-and-forth with Claude/ChatGPT
   Often, step 1 feels too big, so I copy-paste snippets and iterate manually. More friction, but usually faster.

I don’t use Github Copilot, I find autocomplete intrusive.

I feel pretty fast with this setup.
But I’m not “I built a full-blown app while I was in a Zoom call” fast.

Is that claim even real? Can you build an app by typing plain English into a text box?

What does the code under the hood look like? Can you continue from these prototypes, or is it throwaway work?

And ultimately, how’s the actual experience to build like this? Is “Vibe Coding” real?

To find out, I will test-drive the most common AI-powered coding tools. I started with mapping out the landscape to see what’s out there.

Here’s my first pass at a comparison table. Sharing this publicly to get input.

AI Builder Tools

Have you used any of these tools?
Which one should I try first?
What do you want to see tested?