About 15 years ago, with the Icosystem team, I wanted to apply

About 15 years ago, with the Icosystem team, I wanted to apply AI (such as it was back then, at the beginning of the deep learning revolution, the prehistoric dawn of cyborg life it seems) to... IFTTT. I loved IFTTT but I was having a hard time creating more complicated automations for my home (not that I needed them, just bear with me) and I found the programming experience appalling. Great for baby applets, terrible for anything with multiple conditional dependencies. And the other thing I found frustrating was discovery, not just discovery of existing applets that fellow IFTTT users had created, but also, and more interestingly I think, discovery of the undiscovered possible. I wanted to explore possible uses of IFTTT that were not described anywhere, an untouched IFTTT territory that no other human being had ever even thought of. This was perfect for our Hunch Engine, the (interactive evolutionary) search engine for when you don't know what you're looking for but, like Potter Stewart's pornography, you know it when you see it. The team overruled me, they found the problem definition to be narrowly focused on the issues of the out-of-touch tech elite and I did not push back as I didn't have a clear path forward or, worse, not even a business case. Fast forward to 2026: the first problem is solved in principle, just express your needs in English and an AI can assemble the applet for you. It is funny how IFTTT has gotten this backward, helping you create AI agents when what I really want is an AI to help me create applets (an applet vibe coding AI). The second problem is still unsolved: even finding an app you describe in the chat box is hard. Look at the example here. I want an applet that sends me tech news and this is what I get. The more promising capability would be an open-ended tool that allows you to evolve applets without knowing ahead of time where it will lead you.