The Coming Loop

That's the entire topic - loops are not just infinite output, they require automated verification and progress evaluation steps.

The game is to find ways to automate that. Not fully but yes to reduce what's required from humans. Seems like you're questioning the entire premise rather than pondering how far it can be taken and how.

Thank you for writing this thoughtful post, Armin. I find it deeply comforting that the developer of Pi, an agent harness, does not remove himself from the loop, like me. Maybe if I started thinking of codebases as biological organisms I could get comfortable with getting the human out of the loop.

I think what is going to happen is revival of "Methodology".

"Methodology" was a big thing in the past just before we got into "Agile Extreme Coding", instead of trying to model the big picture of SW development projects just jump into coding agilly. Implement it feature-by-feature

Granted the methdologies proposed ( See: https://www.ibm.com/docs/en/rational-soft-arch/9.7.0?topic=m... ) may have been too heavy and not flexible and not improved enough. But now with the rise of Agents I think we need to revise and perhaps re-invent them for AI agentic development.

I think that the problem basically arises due to the fact that the technology is recent and the opinions on it are hard to make and even if we do end up making opinions, then my opinion is of the fact that I am more than happy for AI to suddenly vanish but clearly that can't be the case and there have been some long term discussions that I can point out to.

We can however change our opinions on AI and the culture surrounding it but as witnessed from my comment and other people replying to it, the question is becoming as of if we engineers even have the necessary power capable to bring so the change within how or if we use AI.

That is why, I believe that we might need some nuance rather than a complete acceptation or complete rejection of it.

AI is like a double edged sword, something which is fundamentally both good and bad[0] It is my opinion that currently, the bad outweighs the good and I have been vocal about it, but the control of the matter is strictly beyond one or even one good/bad binary.

The primary issue of the matter is the fact that some people (mainly the C-suite/investor class) believe in so much of the goodness of AI that they are willing to think that anyone pointing out anything bad of AI is the one who hasn't used it and how there are all these anecdotal evidences of people being impressed by model's quality and its evolvement, so surely, if AI must have grown so much and is so good as everyone says online it is (within an internet thought bubble), the fault must be lying on the engineer and he must be one of those luddites who has failed to accept technology and must be replaced.

I suppose some people genuinely believe the above paragraph that I wrote. I have heard some of such sentiments on Hackernews itself...

My point is probably that AI should be treated as a double-edged sword really, that's about it. We can go into the details of it if you so prefer and I think that you are right that its within the nitty gritty details that there is ambiguity and how this discussion leads to that ambiguity, so in that sense I concede but that there should be a better more skeptical discussion regarding the use of AI and it should be up to the organization at first for them to decide how they use AI and be flexible with it but not at the sake of the sanctity of code and coding practices.

I feel like it would be hard to gain a worldwide consensus on the use of AI within software development, we can certainly try but it would be more effective to do so at an organizational level and to slowly level up and I do hope that the world eventually figures out a somewhat consensus of it.

Hope I am able to explain some parts of the severe personality split, this did get long but that's probably because I have just watched some somewhat thought provoking media which used some great vocabulary and I am somewhat copying it as my brain thinks about your comment.

But yea, in all honesty, that's about it. got a bit long so sorry about it but have a nice day and thanks for reading if you did and if not, thats fine as well and I hope you and everyone else including me navigating this somewhat shit hole that has become internet discussion about AI and AI in general, things are getting exhausting and there is some negativity in life at times but I hope that we overcome it and I am optimistic in some sense and I wish nothing but good for all of you within your future as we all navigate the uncertainties of future together :-D

[0]: https://english.stackexchange.com/questions/263521/word-phra...

I've found that if you look over the code and notice and describe a specific problem and solution, the agent can apply a refactoring for you well enough; and that's often faster than editing the file yourself even if you already know exactly what to do.

The idea of setting up an agentic loop to review code and propose and implement refactorings still seems pretty awful to me, though, yeah. Maybe cut that off at the first green-bar revision, and then apply some actual taste and judgement.