Why do people keep coding agents running for so long?
In one firsthand situation, a $100 monthly plan still does not get fully used. The main limit is time, because two jobs leave little room to watch, guide, and review a . There are 3 active projects and about 8 shelved projects, but a typical agent run lasts about an hour at most.
Other people seem to keep setups running almost all day. The practical question is what those agents are actually building or doing, and whether people are even assigning simple tasks like checking email.
Key points
- A $100 plan can still be hard to use fully when human time is limited.
- Having many projects does not automatically create good .
- The main question is what useful work people assign to agents for long runs.
- Agent cost control depends on task preparation, not only model price or .