Speed and complexity; now and then
Many talk about how AI helps speed up development. 10x this, 100x that. But what if the real value of the new technology is not the speed itself, but complexity that is available to us?
Think about it: when you were doing everything by yourself, you had to study & learn the tooling. Then uncover edge cases (often by googling, usually ending up on stack overflow). Then lay out the plan in your head and implement it, often iteratively.
This would take so much time if you were to do everything "properly" that it would be often unreasonable to do it on your own.
Now with AI you can surface above the abstraction—keep your vision and delegate often tricky implementation details to the LLMs. Compiler flags? Project setup? Heck, I am so glad I don't have to pick out of 10 new TypeScript transpilers—each of them "faster" than the other.
I can focus on more complex and broad things. I can also do more proper engineering now: cover things with automated tests, run benchmarks. Actually validate the work.
Sometimes AI code generation tools can be overwhelming by the amount of code they produce, but I feel like with regard to managing complexity they can actually help and improve the workflow.