Is everyone burning money on openai while debugging or its just me?
Im building a saas on my own, using cursor, chatgpt 5 and a small python backend for the agent stuff. the speed is perfect but i’m starting to hate one thing.
every time my agent do something weird (pick the wrong action, retourns empty stuff, gets stuck sticks or similar) i basically have to rerun the whole flow from the scratch to see if my fix worked. each rerun is like 30 sec and a few cents on openai. Yesterday i did it i don’t know how many times, wasted hours and lose 8 dollars just to find a typo in the prompt.
is this just how building an ai agent works in 2026 or I’m missing something? is there a way to not rerun everything every time, or to save the state before the broken step so that I can just iterate on that one piece? im a solo founder, have no fancy team, just me and my cursor..
how the other people here handle this? is it theprice of doing business or is there a smarter way?