hq: An Acceptance Protocol CLI for Multi-Project AI Work
I scrolled back through months of conversations with my coding agents, and one finding stung: I had sent “go on” dozens of times — not because the work was done, but because the agent had stopped. And “review this as an expert” / “try harder” — dozens more. Not because the agents weren’t capable. Because nothing existed that could prove their work. I’m using the best models money can buy. And my management style was: whenever I get a spare minute, check whether they’re slacking off. ...