Hi everyone, I came across this article and thought it might be interesting to bring here: Intent to Infrastructure It talks about the idea that instead of describing how things should be done, we just describe what we want and the platform (with a bit of AI magic) figures out the rest. Sounds promising but also raises a lot of questions. I’m curious: • Has anyone here actually seen this working in practice? • What does it really take to make this reliable in real teams? • Does it actually make things easier, or just shift the complexity somewhere else? I’d love to hear your thoughts. Examples, lessons learned or even cautionary tales are very welcome. Thanks in advance for sharing!

This is a very old concept from 1978

Interesting. Thanks @Piotr Szwed for your detailed response. I'll look into the links you have shared.

And it has to be deterministic 🚔 🚔 🚔

Can you elaborate what you mean by determinstic?

Hey Liora! Article --> TLDR 🙂 The gist seems to be “declarative” instead of “imperative” which is quite an old concept. The problem here, if you throw AI into the mix, is that you can’t predict the outcomes with 100% certainty. So either you establish HITL to check on the outcomes or you follow a different paradigm. Canyon has been a moonshot to enable agentic AI to work together with PE pre-defined golden paths to create the wizard like call & response cycles that drive an AI agent conversation but in the background realize hallucination free and deterministic outcomes. I’m not sure if the recordings are already published but if so, you can get a moderated introduction to the concept in my PC’25 workshop.