the agent tooling is maturing faster than the agents and that is good
the models are roughly where they were a few months back, but tracing, eval, and orchestration got a lot better. the boring infrastructure progress is what makes this usable.
the eval first mindset is underrated, nice to see it here
this is the kind of post I come here for
respectfully I think you are overcomplicating it
thank you for not making this a 20 minute video
respect for actually shipping instead of just theorizing
i was just about to ask this exact question
i was skeptical but the example convinced me
the security side of this genuinely scares me
lol the agent did this to me too and I almost shipped it
how do you keep it from looping forever?
not gonna lie I read this twice and still have questions
skeptical but bookmarking to test friday
i would be careful recommending this to beginners
what does your eval setup look like
we built something close to this, happy to compare notes