This doesn't feel like the right way of thinking about it---if you went from 20% of people paying fares to 80%, that's 4x the revenue with the same fixed costs, and farebox recovery goes from 13% to 52%. Now, I don't think compliance is actually that low, but the point is that you shouldn't think of fares as collecting a fixed share of the operating costs.

The costs may be fixed but so is ~87% of the revenue via tax money.

Muni can be 100% free without increasing taxes too. The people using public land to store their private vehicles are not paying their fair share. Proper parking pricing and enforcement could easily pay for Muni’s deficit.

It is interesting that there's a gap between the "prove N–S existence" conditions, which assume no forcing term, and the "counterexample" conditions, which allow a nonzero forcing term. In theory both (A) and (C) could be true.

It's worth noting that Chinese AI labs aren't profitable, so in some sense they're not doing it without subsidy either.

I think it just might be that this is something that makes sense in some situations and not others.

Yes, I think the idea is that you start with the individual squares, and you can sew two pieces together at a time. If you come to the point where you need to join two pieces along more than one edge, you have an L-seam.

Nonreflexive uses of the verb "pace" are well attested.

https://www.merriam-webster.com/dictionary/pace#dictionary-e... https://en.wiktionary.org/wiki/pace#Verb

I don't think unreadable skills implies proper engineering at all. It's just as or more likely that they're the result of a blind iterative process with no clear improvement signal. (And whether iterative RL over a set of evals is actually proper engineering here is another question...)

> And whether iterative RL over a set of evals is actually proper engineering here is another question...

I'd put it like this: regardless of the merit of how they're applied, it would at least demonstrate possession of the advanced skills expected of experienced software engineers.

Sonnet 5 was released a while before Opus 5, so it's just Haiku that didn't get a 5 release.

Due to their essentially cryptographic nature, I don't think SynthID et al are very easy for LLMs to speak natively. They have other ways of doing steganography.

LLMs with access to tools like code execution don't need channels they can “speak natively".

> Do you accept that the story / justification from the labs in the popular media and political discussion is simply nonsense?

No, and I don't see anyone who's actually demonstrated understanding of what happened in the Hugging Face incident (e.g. reading the reports in their entirety) making this claim.

> Do you believe that any real security was autonomously bypassed without direction by these models during internal evaluation?

Yes. Again, this is hard to deny if you've actually read the reports.

If you were building a sandbox for untrusted code, would you give it access to an artifactory instance outside the sandbox?

I don't see how this is relevant. OpenAI absolutely had sloppy security practices here. But that doesn't mean there was no "real security", and it certainly doesn't mean that their models were simply following orders.

I’m just trying to understand the perspective.

I naturally think of this in terms of running untrusted code. If you really think the agent(s) could get out of control that seems obvious. And the approaches to securing a runtime environment in that situation are pretty standard at this point, and they would never allow something like artifactory access.

And if the question is should they get special government dispensation to form an otherwise illegal cartel, it seems much simpler to just follow standards for running untrusted code.