With IaC rooted so deep into infra management with its declarative and deterministic behavior, using agentic CLI tool that is never deterministic feels quite unnatural now.

Oh thanks for watching! 💚 In short words, it’s a combination of support from leadership and real value for the teams (by co-creating templates with devs). Of course not 100% use them but we “softly” close the possibilities of custom setups making them exceptional. So majority accepted it well.

Should we expect a podcast? This would be an amazing duet!

Well, there is an AI CEO as well; I bet we can gather a full staff with those tools so they could launch a startup 😀

So far, the one perk I know to improve the decision-making capabilities — is to use the sequential thinking MCP from Anthropic. This adds up to the total “thinking” time but results are generally better and decision making is more reliable.
github.com/modelcontext...

Nice article; thanks for sharing! That one about “blind copy-paste coding” is a good sign indeed, but it kind of contradicts those agentic features when an assistant can make code changes for you. Anyway, I believe over-reliance on AI makes one a weaker engineer in the long run.

IMO, there is nothing shocking in that as long as it is the human then who reads that code and commits/pushes to git. I write a lot of code with AI but I read and adjust that code then. AI coding is a powerful instrument but it does not mean that we remove all the guardrails we have.

Wonder how it could do the changes described (it’s clear that it did modify TF files code) and at the same time had issues with other tf files. 😀

😄 Not only at large ones. What I observed, though, is that people magically find the ways to contribute if they *really* need the change. If not, they are OK to wait. Or even to abandon the idea. But the outmost blocker is docs or large code base which is less of an issue nowadays with AI tools.

I also see how it gains popularity in secops tooling too. But then, of course, this ML > AI popularity evolution is the most prominent and well deserved. Wonder if Go will get some boost besides cloud-native solutions and utilities-for-any-kind.

Sunday is good for cats, not for releases :-D

You’re now now just one step away from using some automation for these queries to send you an alert if the response crosses some threshold. MCP-SRE basically 😄

Seems like instead of focusing on the genuine customer pain and making a real value and the purpose of the apps from that, a lot of companies are adding AI as a “magic solution to everything” and putting that fact on a spotlight with lower focus on discovering real customer pains or needs.

Some punch about JavaScript vibes and Electron-ish ubiquitous could be here, but I’ll just say that I feel your pain 😅