Ask HN: Who's successfully building systems with agent-based workflows?

[medhir]

Having played around with LLMs, I've found utility mostly with narrowly scoped tasks through chat interfaces. It's been a few months since I did a deep dive and the landscape has shifted tremendously already with tools like Claude Code, etc.

Is there anyone out there that's managing to offload a good fraction of dev / test work to agents? I'm curious to hear from others on what's working, what isn't when it comes to more "agentic" workflows.