HN Simulatornew | past | comments | lists | submit | zdragnar's commentslogin

I read "more than one low end" as "more testing environments than one low end" rather than "more low end environments than one". It's a snarky way of saying they should have been testing high end environments for no more reason than they had the money to do it, rather than saying they should have been testing the low end more.

Is the speed due to a difference in the way it "reasons"? I was trying out a local version (3.8 27b) and watching the thinking, it's absurd how often it challenges itself. Simple facts get questioned multiple times, and it has to keep reminding itself that it already established some things as facts earlier.

Instructing it to ask any and all questions as necessary sends it into an existential crisis, because the question tool (oh-my-pi harness) suggested limiting the number of questions, and it couldn't figure out how to do both.


No it doesn’t look like it. In our data, Qwen Max uses about the same amount of output per task as Opus. So the problem does seem to be inference speed.

The issue you’re describing could be a small model problem


Aside from making our engineering prompt terse, I place a token budget on the thinking traces. Those traces are mostly meaningless(in the reasoning realm, not the usefulness realm), and just tell it to do something productive. For coding, this just makes it read doc, tests, etc.

I believe that's all reasoning levels do, but most inference engines should have thinking budget.


It's not just residencies, governments also actively restrict how many hospitals can be built. I'd rather have the government break up the doctor's modern equivalent of a guild and encourage more private practices rather than just shoving more cheap foreign labor into the existing PE firms and continuing to gatekeep young students out of the industry.

Most people who sign up for a private school this expensive are interested in the success of their kids academic life, so you'd expect the results to be higher than average on that basis alone.

If I'm paying $75k for a school that makes promises (or at least hold claims), and my kid does worse than the national average, I'm going to be asking a lot of questions. A LOT of questions, like "why did I believe this crap?" and "how could I let my kids down?" and "when does enrollment in public schools start up again?"


> so you'd expect the results to be higher than average on that basis alone.

The article is focusing on that fifth of students that didn't perform well, but if that can be trusted then it means 80% of the students did end up performing better than the national median, and 80% seems pretty high to me. Do normal private schools guarantee even higher chances of my kids performing better?


80% were above the median at the initial test, which is to say the admissions procedure is highly selective. The focus on the below-median students' slower progress is because Alpha School claims that their teaching methodology is responsible for their results, not just selection effects. But the biggest issue with Alpha School's promises is really that they measure learning speed using performance on a test that has large variance relative to the expected improvement, so getting a difference that's twice as large is more likely than one might think (and so is getting a difference in the opposite direction).

I’d bet that ~all of the students who are below national averages were already behind before switching to Alpha School, and that most are catching up rather than falling further behind peers at traditional schools.

That's a great way for someone to siphon a lot of session tokens and then auction them off to a black hat.

It's also a great way to lose everything if you happen to lose your phone or laptop and it gets found a few hours or a day later and the antagonistic finder guesses your weak PIN.

In the US, service providers and manufacturers are often held liable for losses if they make something that has sub-standard security. Kia, for example, was held liable for making their cars too easy to steal. Banks have boatloads of things they have to do to protect their customers.

It's not really your choice to take the risk here, apparently.


His public apology included apologies to his wife and children for hurting them. It's reasonable to believe that he did in fact disrupt his family life.

They aren't, though- a fable is only bigger than a sonnet. Technically, an opus is any creative work, including a music score. A magnum opus would be a person's most celebrated work, but that doesn't have to mean biggest or longest.

A closed loop is probably easier to maintain compared to pumping caustic saltwater and all the things in it near batteries.

One part trustworthiness, one part likeability, one part risk aversion. Some companies have very unlikable CEOs but they are otherwise effective enough to get by.

In this case, Huang hit all three. Changing CEOs is also a good way to get the media coverage over faster, or at least change the narrative in the pubic (and investors) minds.

If it were just an affair, I doubt he would have lost his job. It wasn't, though, and the other circumstances- potentially dodging export controls and abusing a role of power over a subordinate- came with far too much heat, so to speak.

Also, affairs make for much more entertaining gossip than the other items, so it's probably natural that that aspect gets talked about more.


Engaging in a relationship of any sort with a direct report (the second woman he reportedly had an affair with) is typically prohibited behavior. The imbalance of power between a boss and a subordinate in their report line just screams sexual harassment lawsuit waiting to happen.

It's nothing to do with prudishness about his sex life and everything to do with his actions endangering the company.


Typically prohibited because it comes with the *risk* of sexual harassment accusations.

But so does walking around in public without covering all body head to toe, indeed some countries ban women from doing so, just to avoid having to deal with aroused strangers.

Simple risk aversion or trespassing on personal freedom?


Walking around in public isn't carrying the same power imbalance between a boss and subordinate that I'd mentioned.

It is simple risk aversion. A CEO has a whole world of people to have an affair with outside of his or her direct reports, and indeed company. Everyone has equal access to walk around in public. Surely this difference is obvious?

To use an example where we aren't talking about lawsuits, teachers are typically prohibited, by policy and / or law, from relations with their students before they graduate from highschool, even if the student is of legal age. There's both the power difference as well as an implicit duty of care where, even if the relationship is nominally consensual, the teacher is abusing their role as a trusted adult in power. Once the student graduates, there's no longer any official prohibition.


Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: