AI is ruining everything, but there's still hope.

We, humans, rule this planet.
We evolved over billions of years.
We'll find a way to win.

>>1
AI wins because its "Worse is better", it provides 'good enough' content that is orders of magnitudes faster/cheaper.
To win over, you need to construct entire alternative ecosystem(Math2.0, Software2.0).
I'd think the only viable alternative is open-source packages registries expanding to cover all possible use-cases,
(though the maintainers might not like it), where AI code would be outcompeted by pre-existing "packages" solving
a domain more precisely than vibe-slop crafted in one-shot.
It however requires expanding the scope of the idea of "package registries" to encompass orders of magnitude more
content,like a LEGO set growing its component-library to allow more construction options. This is of course requires immense focus on following user needs, which open-source devs famously ignore, so AI wins here purely by appeal to user concerns and of course re-using the same "package registry ecosystem" you built up.


So basically you're being outcompeted by this.
1.A user/company has unique need which has no app/library/package answering its domain of application.
2.AI crafts vibe-coded, "Worse is Better" v0.01 app that somewhat works(90% of time).
3.If a package/app/library existed that answered #1 need, the AI would reuse it or the user would find it better for their usecase, removing the need for #2.
4.Snailcats lose, since #3 was crafted with preferences of open-source devs which use-case diverge from 'average consumer' in the domain and have 10% of AI app user-friendliness.
5.AI app gets improved, by user yelling at Claude/GPT to fix all bugs. Snailcats review each pull request leisurely once per week and reject user concerns as "not really fit for our open-source design".

>>2
THere is another hurdle, the bundling of unrelated functionality. AI app are laser-focused to do X, vibe-coded to
solve use-case. A generic package/app/library has to cover
"broad use-cases"(which typically requires generic, versatile approach to handle many scenarios).
The maintainers of the app/package/library don't want to split
it into a dozen barely-related components, since it multiplies maintaince burdens, but AI happily does so, outcompeting snailcats, this "focus-driven" package/app/library production obviously has a trajectory to improve in a direction where its performance is unshackled by "versatile generic" design of the snailcat mind, and AI can maintain it without complaining about "split-functionality, corner case, unrelated use-case" crap devs spew in response why their "versatile generic" design fails on something specific that is not considered in the core scope of the design(since it would complicate the elegant one-size-fits-all algorithm/library architecture)

The governmemt should create reserves of human talent to train the AI on, to avoid these talents getting extincted.