What would you let an AI agent do if you could rewind it?

Hey everyone, I'm Drew, building kosa8. We launch here on Tuesday 20 October.

The question behind it: AI coding agents now run with your keys, your data and a shell. Most of us either babysit them or give them less to do than they could handle, because one wrong command or one poisoned web page is all it takes.

kosa8 gives every agent its own sealed machine (a real microVM), with the network closed unless you name a host, risky actions waiting for a person, and checkpoint and rewind of the whole machine in about a second.

Before launch, I'd love to hear from people actually using agents:

1. What's the one task you don't hand to an agent today because of risk?

2. Has an agent ever broken something for you? What happened?

3. What would make you trust an agent with more?

I'll read every reply, and the answers will shape what we show on launch day.