No, it’s not about me not understanding how to test the behavior, it’s about me not understanding how to test that behavior given the current state of the codebase (think Working with legacy code)
Sure. I've alluded to your point about "outsourcing code maintenance to AI, so we could buy ourselves more brain cycles and widen our bandwidth to focus more on the what part of the equation". If only leave ourselves the "what" part we may run into a situation when forward progress is very hard or even impossible without large-scale refactoring (or even a complete rewrite).
Thank you, Andre. That's a great question. In my experience, the speed of development is never the bottleneck in the SDLC. The new development model, powered by AI. will hopefully bust that myth, so that we can abandon the false hope that speeding up the development process is where the biggest gains lie. Then, we'll focus on the real bottlenecks, which are mostly caused by working in secrecy.
I believe the AI-powered SDLC will make teamwork much more visible, and by virtue of being visible, much more value-focused.
> I want to see if that process will lead us to the point where AI starts losing its ability to make the next test pass.
Or you start to lose the ability to write next test or comprehend the system. The former could be a consequence of the latter)
How do you lose the ability to write the next test? If you lose that ability, it means you don't know what you are building.
No, it’s not about me not understanding how to test the behavior, it’s about me not understanding how to test that behavior given the current state of the codebase (think Working with legacy code)
That's a common situation and has nothing to do with augmented coding using AI.
Sure. I've alluded to your point about "outsourcing code maintenance to AI, so we could buy ourselves more brain cycles and widen our bandwidth to focus more on the what part of the equation". If only leave ourselves the "what" part we may run into a situation when forward progress is very hard or even impossible without large-scale refactoring (or even a complete rewrite).
Yes. Obviously, such an approach (i.e., outsourcing maintenance to the machines) only makes sense if 100% of the code was created by the machines.
Great article, Alex. It will be interesting to see how the traditional SDLC will shift to the new development model.
Thank you, Andre. That's a great question. In my experience, the speed of development is never the bottleneck in the SDLC. The new development model, powered by AI. will hopefully bust that myth, so that we can abandon the false hope that speeding up the development process is where the biggest gains lie. Then, we'll focus on the real bottlenecks, which are mostly caused by working in secrecy.
I believe the AI-powered SDLC will make teamwork much more visible, and by virtue of being visible, much more value-focused.