Dani from Web Reactiva had me on his podcast to talk about how I use AI day to day with Drupal. Not the theory, but what I actually put into projects with paying clients and code that ships to production.
We recorded it, he summed it up in his newsletter, and he put it together really well. He sends the newsletter through Substack, and the episode is embedded right there, ready to hit play:
Multiplícate con la IA (contado por un freelance)
Heads up: both the newsletter and the episode are in Spanish. If you'd rather listen in your usual app, it's also on any podcatcher (Spotify, Apple Podcasts, whatever you use), just search for Web Reactiva.
What we talked about
A few of these are things I've already covered here, and a couple I haven't written about.
We talked about why AI doesn't commit on its own in my projects, how I push back on it by handing the code to a different model so it can poke holes in it, the Ralph loops for leaving tasks running overnight, setting up a system with several models in a hierarchy instead of always reaching for the most expensive one, why I dictate instead of writing two-line prompts, and isolating everything in Docker with DDEV.
The part about not trusting AI blindly and reviewing everything I covered with concrete examples in six real-world AI cases in Drupal this week. Ralph, container isolation and the agent system are all in the article about my full Drupal and AI development setup in DDEV. Not always reaching for the expensive model ties into how to optimise token usage in Claude Code. And the dictating part I explained in depth in dictating to AI on Linux, no internet required. If you want the broad picture of where I fit AI into Drupal, that's in how to use AI in Drupal for developers.
Plenty more came up in the episode beyond these, and on some of the topics I've already touched here I get into more personal takes or more detail than what I put down in writing. So if any of this interests you, I'd recommend giving it a listen. It's worth it.
Subscribe to Dani's newsletter
Beyond me showing up in this issue, I'd recommend subscribing to Dani's newsletter. It's free, he covers genuinely interesting stuff, and lately he's been deep into AI applied to programming, which is right where I am these days.
If you listen and spot something I'm doing wrong, tell me. This changes every few weeks and I'm still tuning it.