AI made code production cheap and fast. The three feedback loops that turn code into validated value still run at their old speed. Managing that gap is what mature AI product development now means.
I'd like to add one more aspect, that I didn't read: The review in breakpoint one is not just stressed by the significantly raised volume. It's also stressed by the no longer exiting evaluation-while-writing that humans did. That typically was the first, and maybe crucial, step of answering the question: "Does this make sense?" - which was applied keyword-by-keyword, line-by-line, function-by-function, commit-by-commit on different levels of abstraction, even before any PR - or whatever process triggered a review - was submitted. The absence of it may be a even larger issue than the increased volume.
Yes, you’re right. It’s a consequence of what I wrote on option storming. But then it’s still a question if everything on that level is always on the same CI loop orbits even reviews rather on demo / prototype level in more sloppy quality.
That would still be a choice per org and per feature / product etc.
But by standard you are 100% right. Thanks for pointing that out.
Great article.
While the diagnosis is correct, I think that you are missing one thing:
Time orientation. It's the only way to get to actual flow, in software development (and elsewhere: https://betacodex.org/white-papers/paper/introducing-time-oriented-software-development-26
Here's an article about how this connects with AI-aided development and the "loops" you mentioned. https://nielspflaeging.substack.com/p/ai-aided-software-development-needs
Hi Niels, thx! I‘ll look into it and reply. Thanks for the hint.
Love it.
I'd like to add one more aspect, that I didn't read: The review in breakpoint one is not just stressed by the significantly raised volume. It's also stressed by the no longer exiting evaluation-while-writing that humans did. That typically was the first, and maybe crucial, step of answering the question: "Does this make sense?" - which was applied keyword-by-keyword, line-by-line, function-by-function, commit-by-commit on different levels of abstraction, even before any PR - or whatever process triggered a review - was submitted. The absence of it may be a even larger issue than the increased volume.
Yes, you’re right. It’s a consequence of what I wrote on option storming. But then it’s still a question if everything on that level is always on the same CI loop orbits even reviews rather on demo / prototype level in more sloppy quality.
That would still be a choice per org and per feature / product etc.
But by standard you are 100% right. Thanks for pointing that out.