Where's the boat?

If you're a maintainer, then I suspect whatever you maintain fucking sucks.

I clearly explained the problem- that the quality of LLM-produced output is a function of the money spent to produce it (generally).
This means that the flood of slop PRs are produced by free-to-cheap models. This is only natural.
People who have no idea what they're doing aren't spending $300 for a large commit. They couldn't be bothered to invest the time in learning to code, they sure as fuck aren't going to invest in this.

Meanwhile, over here in real-life, dealing with real problems on maintained things that don't suck, I've got people making quality commits with LLMs on the daily. But they're part of the team, and they're paying real money for those commits.

In short, I think you're probably just full of shit. If you're not, I feel for whatever the fuck it is you maintain. You can't even apply basic logic to a problem.

As someone who is dealing with this problem, it's not that simple.

The problem with your assessment, is that there is no "the LLM".
LLMs come in a fucking vast spectrum of capabilities.
Put pretty simply, very expensive models do very good.
After very expensive, you get a spectrum from "pretty damn alright", to "outright terrible."

There is no "the LLM."
I can tell from your comment that you're talking out of your ass, and not actually dealing with this problem.

> It's a nice idea, but who equips their trailer with this to save money for the stranger hauling it around?

Fleet operators, where they don't have strangers hauling around their trailers; e.g. USPS, Amazon, FedEx, UPS.

=Smidge=

> Once, the one time 5% is spent the state will have to figure out how to do the one time 5% more than once to keep feeding the spending machine.

The money to be raised is already budgeted as a separate fund for a specific purpose. It is not intended to be general funds nor is it intended to be ongoing funding.

Think of as a bond. When the government wants to raise money for a project or investment in the future, they will often issue and sell bonds to raise that money. Bonds mature and pay back with some interest, and are not recurring or factored into the normal budgeting.

This is functionally the same thing, except instead of borrowing via bonds and paying back with interest, it's just a straight up tax on billionaires.

=Smidge=

Not sure if it's an age thing or what, but you're not supposed to take it as a given that the police can set up a sting based on automated output from an unaccountable machine; even if the police themselves were perfectly professional about it.

The issue is not whether the cops acted professionally. The issue is that, even if the surface level outcomes seem similar, there is a huge difference in your actual level of freedom in a world where a someone has to convince a judge that your behavior is suspicious verses one in which a bunch of automated and bureaucratic systems can flag you as a problem with no means for you to make them account for their reasoning.

This is a projection of personal issues.

Normal people find "not being able to comprehend other person's words" to be a much greater obstacle to comprehension than "psychological issues".

Because most people are sane and have a reasonable amount of control over their psychological issues.

For totalitarians, need to identify people is critical. It's why DPRK is indeed democratic and holds elections where Kim wins every single time.

Because voters are identifiable.

This is why 16 gig slow GPUs are more expensive right now than very fast 12 gig ones. More RAM means bigger model can be fit into memory. Also why 24 gig models are unobtainium.

4090s and 5090s have been long used for narrow models in things like research in much of third world. This is the natural next step. Shrinking models further so they can be operated from phones that already have good enough cameras to enable things like discriminating vision, where model helps identify specific things that camera looks at. We're looking at the next big ag breakthrough in marginal places like Sahel. We now have an actually good chance of AI machine vision doing something we just couldn't do with mere human vision and algorithmic machine vision. Identification of pests and weeds, their eggs and larvae, etc.

And in medium to long term, it looks even better. We may have an actual viable chance of eradicating Tsetse in medium term future with AI vision. One of the main reasons why Sub-Saharan Africa is still absolutely fucked in terms of human development may actually be finally removed.

X's "automatically translate into your language" feature has actually rebuilt the tower of Babel. The main barrier in interpersonal communication came crashing down.

I suspect that minors can to an extent form contracts. For example, to buy anything from a store, you have to fulfill a contract (money for product). This does not require parental intervention. They can take on a summer job. Same thing.

Parental intervention only comes when there are contractual violations (i.e. minor shoplifts, parents are on the hook for at least some if not all of the damages). Though in some cases/places, parental acceptance can be required for taking on a job as a minor.

So the argument they're making is most certainly legally sound in this aspect. The point of contention is going to be "how much can a minor contractually agree to on his/her own" and "with this much additional obligations, how can amount of obligations for each contract be verified".

The whole thing is a horrible mess because before ease of access granted by computerization and proliferation of networking of computers, most contractual obligations required some kind of face to face verification of basics. When you applied for a summer job as a teen, you got into an interview where your identity got verified.

Meanwhile today, the issue is that due to aforementioned factors, we now have countless contracts that are made without parties ever encountering each other in a face to face situation. This seems to be, on a fundamental level, an attempt to get something that is as close to "face to face" as can be generated in age of social media. I.e. other party gets to at least verify if you're specially protected kind of a human (minor), or one that is fully legally responsible for him/herself. Something that used to be done face to face for each contract that extended significant legal obligations for both parties.

But it's clumsy to the extreme, which is likely the main reason why it's being fought over. Essentially instead of "protecting the companies from having unknowingly taken on additional obligation of contracting with a minor", which the legislation purportedly is trying to do, it instead just punishes everyone who isn't a minor (both users and companies) by forcing companies to assume that unless proven otherwise, everyone has those additional obligations of contracting with a minor.

Find an independent EV specialist. They can probably get or build you an aftermarket LEAF pack with more modern guts.

There's a ton of aftermarket support out there.
=Smidge=

> Might be. But it is not far outside the window of actual expense

"Yeah the evidence is fake but that doesn't mean the conclusion is wrong."

That's basically politics in 2026 in a nutshell, I guess.
=Smidge=

Aside from the early LEAF packs being notoriously bad with degradation - both due to early tech AND bad thermal design - it's also worth noting that the main reasons EV batteries enter the secondary market is because the vehicle they were installed in got totaled.

=Smidge=

The most important difference in all your examples is things like tooling and equipment are either strictly necessary to complete the job, or produce so much value in terms of productivity that they are worth the cost. Also, most tooling and equipment lasts longer than an AI token so the cost tends to get spread our over several jobs...

Using AI coding agents has not proven to increase quality or productivity in any meaningful way - increased volume of code does not mean productivity unless you're a middle manager. It's known that it is not strictly needed for the software engineer to do their job. You are not improving the engineer's workflow by mandating AI use, you're just making it more expensive.
=Smidge=