I think it's even more pervasive than that. Why bother forking software at all? At the point in which code generation is meaningfully trivialized, software becomes entirely disposable. Anything you want, have a model spin it up. You don't even need libraries, the model can just make everything in-situ, who cares? Why on earth would I ever want to use SQLite if I have access to a sufficiently advanced code generator which can generate me a similarly high quality database system, with the added benefit of conforming to whatever my problem domain is, conforming to whatever branch of database theory I want?
Even SaaS isn't safe. I don't even have to describe your product to my system, I just have to give it a harness with access to the interface and have it replicate it locally. Frankly you can probably already prompt for that.
The only thing holding this future back right now are pricing problems and code generation quality. Both of those barriers are constantly being knocked down. We might never arrive at that future, but it's definitely a higher probability than solving AGI's scaling issues, and would arrive much sooner for technical users.
> Why on earth would I ever want to use SQLite if I have access to a sufficiently advanced code generator which can generate me a similarly high quality database system
Because SQLite has 10k requirements that wouldn't even cross your mind to write down, but 80% of which are useful to you.
The nice thing about natural language is that nesting semantic layers is free and arbitrary, and far more tractable than in a formal grammar. Every natural language is like coherentist ω-order logic. Effectively, I don't have to write the 10k requirements. I only need to provide a sufficient metatheory that can be extrapolatable to those 10k requirements, and that can include embedded theory I did not write myself but am familiar with enough to invoke, as well as refinement criteria ranging from the fuzzy to the explicit with priority weighting parameters to describe the shape in which I want the search space pruned.
This isn't anything new or particularly interesting. It's the entire basis upon which ILP demonstrated generality. A metatheory to synthesize 10 trillion rules isn't even scratching the surface of what you can reasonably do. The key was finding out the tractable semantics for actually computing it in reasonable amount of time, which right now is looking decidedly like informal semantics was the answer the whole time.
Not sure what this is a parody of, but it's hilarious, thank you. Keep it up!
I accept your concession. The humiliation of irrationally committed foundationalists has been a long time coming, so it's good you're trying to get ahead of the curve.
>Why on earth would I ever want to use SQLite if I have access to a sufficiently advanced code generator which can generate me a similarly high quality database system
https://www.youtube.com/watch?v=V_qzqY1bb7I your sufficiently advanced code generator may generate you a high quality database system for some measure of quality, but it will not have SQLite's reliability over the extremely long tail of edge cases proven through its testing and use in real life
Then you've failed the criteria of sufficiently advanced. It's perfectly fine to cast doubt we'll see scaling to this generalization, but you're not casting doubt you're outright rejecting the premise in-confidence. It betrays that you have no idea what you're talking about. May I see your quantification of this long tail? Do you even know how to formalize the mapping from n-bit precision of weights and/or activations to the standard deviation of a transformer's output distribution, such that we could decide whether the long tail of a given behavior is unreachable? Something tells me that no, you don't know how to do that in the slightest. So what drives you to speak with such confidence?
That's before we get into the entire non-linearity of agentic systems introducing massive decidability problems on this in the first place. A little bit of epistemic humility please.
You still spend all your time troubleshooting the reinvented wheels even if AI writes it because you won’t know the edge cases til you hit them. Then you modify the lib, re-release, and update all your apps, but now look where all your time is spent.
The assumption you make is the classic LLM mistake of thinking writing code === building software.
To cite your example SQL has had its tires kicked a lot it’s seen things you can’t even imagine thanks to being used millions of times by millions of people. It’s hard to just replicate all that iteration, learning, mastery, and process. If you reinvent it, users will encounter the dumbest bugs over and over and over. Sure you’ll fix them, but you’re now embarking on this big thing that SQL and others already did.
If you love the problem space definitely do it - go full steam ahead - especially if you’re actually innovating and doing things better, but don’t be fooled into thinking anyone can, or should with every side project.
The new struggle is focus, what not to build, I almost have the purely opposite view of instead of using LLMs for grandiosity, only using the LLM for tedium and making sure it doesn’t do anything too much that I haven’t planned for or want to do. I drive the thing, so every new project is still my time and energy and focus.