something as simple as a wrong data structure used or the wrong pattern can cost thousands of lines of code down stream
This, in turn, can cost hundreds of thousands of more tokens spent, per problem.
This, in turn, will be replicated more and more as the official pattern and your code, though fully maintained by an agent will cost more to maintain.
Just pure software problem, picking the right abstractions is hard.
This is also why I think webdev has seen such a success. React has 1 billion examples of "the react way of thinking" and thus the model can reason more correctly for some set of software. The amount of data structures and their implications are rather small when it comes to a lot of basic web dev. This means the cost / angst people have about models, I think, are often loss on the less experienced because their experience has demanded one dimensional structures.
It almost reminds me of economies of scale when it comes to the physical world. Small 144th RC drift cars spin like wild and are fun to drive, but that is because their square relationships are small enough that it almost aligns with linear relationships. But as the car grows in size, the squirliness goes away because of the same squared relationships.
yall im ngl its way harder to get joy and satisfaction out of building with ai than it was before
constantly straddling burn out, being far less immersed in hard problems, constant context switching
i’m tired of the uncertainty of where this is going and how to do it well