if you want self-improving software, you need to fit your app in a context graph
that's the only way to enable agents to understand everything you run
I personally built this over the past 6 months, here are all the learnings, feel free to copy
https://t.co/uFkhstkASg
if you're building sub-agents you should stop now
our main agent used to be a workflow with sub-agents, each with a specific role (triaging, investigating, etc.)
we moved to a single agent and latency dropped, quality improved and cost was slashed
https://t.co/2UzloDY5lC