Comment by throwaway219450
1 month ago
The agents always do the best work IMO if you already know exactly what you want, but are too lazy to implement it. I like having the agent mock up a working solution before reimplementing it.
To split the difference, I now try to hand code as much as I can from the beginning, leave TODO comments for the agent to mop up and I'll ask it to complete the issue with reference to the current diff. It reduces the surface for agents to make stupid assumptions. If I can get it done fast on my own, win for me, if the agent finds issues or there's logic that needs checking, also a win. This way you stay sharp, but you have access to an oracle if you get stuck and it costs you fewer tokens.
Yeah, I like the "get out of jail free" card approach. The thing I always used to hate before this era was getting stuck in a hole on something that would take days or worse to grind through. It's nice to drop a little plank bridge across those now