Howar Sullivand

What breaks when your AI agent needs a human answer?

I have been seeing the same pattern across coding agents and desktop agents: the agent can do more of the work now, but the flow still depends on small human approvals at the worst possible moments.

A shell command needs permission. A file change needs review. A calendar action feels too risky to run automatically. A tool call looks safe in one repo but dangerous in another. The agent pauses, the user is away, and the whole run loses momentum.

I think AdSense rejection emails are one of the least helpful feedback in the tech field. They simply state, "Insufficient content." "Policy violation." They don't even specify the exact page, section, or rule.

When I spoke with bloggers who had this experience, I found the pattern was almost always the same. They guessed the reason, finished writing the post, waited two weeks, and still got rejected again. Some people even gave up after trying three or four times.

Has anyone here had a similar experience?

Did you actually figure out the reason for the rejection?

A big part of my week is spent talking to early-stage founders about how they handle login, access, and identity, and there is a pattern I keep noticing that I want to understand better.

Almost everyone puts auth off at the start, which honestly makes sense when you are still trying to prove the product matters at all.