kevingadd

One potential virtuous cycle here is that accessibility trees used by tools like screen readers are also a nice potential way for a model to consume information about what's on screen and how it can be interacted with. So it creates an additional incentive for improving the accessibility of new and existing software, because doing that lights up integration with future models.

Quoting the start of the paper:

While MCGs and LCGs have some known defects, they can be used in combination with other pseudorandom number generators (PRNGs) or passed through some output function that might lessen such defects. Due to their speed and simplicity, as well as a substantial accrued body of mathematical analysis, they have been for a long time the PRNGs of choice in programming languages.

EDIT: And going further, they call out Marsaglia's work in particular, it seems.

Since the 1980s, parents have grown more and more afraid that unsupervised time will expose their kids to physical or emotional harm. In another recent Harris Poll, we asked parents what they thought would happen if two 10-year-olds played in a local park without adults around. Sixty percent thought the children would likely get injured. Half thought they would likely get abducted.

These intuitions don’t even begin to resemble reality. According to Warwick Cairns, the author of How to Live Dangerously, kidnapping in the United States is so rare that a child would have to be outside unsupervised for, on average, 750,000 years before being snatched by a stranger.

I wonder how we ended up in a situation where people think Stranger Danger is this bad. Is it just from TV and the internet inflating the danger to drive views/clicks?

In many areas crime has been trending down but people seem to think things are more dangerous than ever, in general. It baffles me.

Just to be clear you're literally repeating a conspiracy theory. The thing you described effectively does not happen. We can talk about whether it should be allowed to happen, but at present it doesn't. Policies followed by most medical professionals simply don't allow it.

Where I live in Seattle we fund keeping the streets in good condition. I see city staff roaming around during the day from time to time wearing hi-vis, doing stuff like picking up trash or removing graffiti.

If trash is lying around only getting picked up by generous citizens in their spare time, what that implies is that the city/county have chosen not to invest in maintaining the streets, and the citizens have elected to throw trash everywhere. I don't think we should take either of those conditions as a given. Better things are possible.

I see white on dark purple at a perfectly legible size using a regular font. Did an extension you have installed block loading of an image or style sheet?

Had to buy an IPv4 address for a VPS the other day in order to clone some git repositories. Couldn't believe it. Costing their customers money when they should be able to support v6 by now.

Can't you already do this by creating a tree of invisible DOM elements with the right ARIA attributes?

In a judgment that spanned 2000 pages, Australian Federal Court Justice Jonathan Beach, ruled that Apple had a substantial degree of market power. The Judge said both Apple and Google had breached Section 46 of Australia’s Competition Act. The companies had abused their market power to stifle the competition. But, it wasn't all in favor of Epic Games. Beach rejected the claim that Apple and Google had breached consumer law, he also said that the companies had not engaged in unconscionable conduct.

2000 pages! I can see why the case took something like 5 years.

It sounds like a mixed ruling so Epic didn't get everything they wanted here, but if they're able to launch the Epic Games Store on iOS in Australia that's a pretty big win by itself.

Typically when a human has a disorder or limitation they adapt to it by developing coping strategies or making use of tools and environmental changes to compensate. Maybe they expect a true reasoning model to be able to do the same thing?

Almost every mention I've seen of gpt-oss was a complaint that the training on synthetic datasets produced a model that's mostly good at benchmarks. Are benchmarks the great results you're referring to or are there a lot of satisfied users out there that just don't post here on HN? Genuinely curious.

I can see how performing well on benchmarks at the expense of everything else counts as great results if that's the point of the model.

Replies seem to be arguing that this wasn't a 51% attack and was something else. I don't know crypto well enough to verify their claims, though.

The country's money is his money. For example: That plane that got 'gifted' to him is getting expensive retrofits on the taxpayer's dime, and he gets to keep the plane after he leaves office.

https://www.nytimes.com/2025/07/27/us/politics/air-force-one...

Part of the complexity here is the temporary object allocation and the need to read properties out of it as you iterate. This is one of the things I tried to push back on but the spec people were in a hard spot, since JS doesn't have native multiple return values. I think in ideal circumstances the allocation can be optimized out by doing store-to-load forwarding on the properties, but in cases like the post here since inlining can't happen, no such luck. :(

Depending on where you work they might not let you publish a paper about it. Certainly was the case at one game studio I worked for, very secretive.

If you're using a portable library that needs to render graphics on mac, it's probably using OpenGL to do it unless it has a platform-specific backend.

This article feels like it generalizes quite a lot about a very, very large population. I'm always wary of pieces that try to put the vast majority of people from a given territory into some broad buckets in order to draw conclusions about them. There are a grand total of two citations in the whole thing unless I missed some, and they're not supporting most of the generalities.

Much of East Asian culture is rooted in the teachings of Confucius.

Really?

Interesting topic at the very least. Any time there's a big population disparity in selective groups it's usually worth investigating how that disparity came to be to see if we can learn something from it and potentially equalize it (if it exists for a bad reason).

It usually happens by accident. For example, let's say you have a class Person with two data members named firstName and lastName. We're already in trouble, but let's make it worse: It has a getter named fullName which returns $`{this.firstName} {this.lastName}`.

That getter looks inoffensive and will, depending on your requirements, work just fine. But it has side effects because the string interpolation allocates and could trigger a garbage collection.

Note that if you're using modern JS 'class' blocks a 'get x ()' will be ignored by JSON.stringify, so if you're aiming to reproduce this you have to use old-school Object.defineProperty instead.

Calling a property getter can have side effects, so if you serialize an object with a getter you have to be very cautious to make sure nothing weird happens underneath you during serialization.

People have exploited this sort of side effect to get bug bounties before via type confusion attacks, iirc.

I agree that it is probably occurring at multiple points in the pipeline. We recently observed one attempt at fixing this earlier in the pipeline (affirmative action at universities) being struck down, though - perhaps for good reason. So it seems like it's not allowed to fix it at the university level, and it's not FAANG's job to fix it in their hiring pipeline, so whose job is it to fix it? Elementary school teachers?

My instinct is that it has to be addressed broadly across all stages of the pipeline for there to be any results.

I would expect you can only transmit so fast in morse code given the need for the dots and dashes to be clearly distinguished from each other and identifiable by the recipient.

Of course, if you know both ends are computers you can just transmit in some other encoding at a much higher rate.

The fundamental question is whether you think the population distribution of a company with thousands (or tens of thousands) of employees should look similar to the population distribution of the countries it's based in.

If the distribution doesn't look similar, why not? Is there a good excuse for being predominantly male, or predominantly white, or predominantly asian, or predominantly indian? What does that good excuse look like? I see people complain about these things all the time.

There's an assumption here that discrimination in the workplace only exists because of DEI which is a weird leap to make IMO. You could argue that DEI made the situation worse and that gets us somewhere but it's not clear to me whether that's your position.

Isn't that $4b in raw revenue per month, or $48b/year? That's not even considering the expenses to operate the service and deliver the $200/mo product. I don't see how that automatically equates to a $300b valuation. A high valuation, to be sure, but...

I'm still very curious whether OpenAI is turning a profit on any of their services.

Is their name particularly relevant to the headline? If anything it feels like it might be beneficial to emphasize that it's not about the who here, but instead the what.

Maybe this is just a fundamental moral divide but a lot of people do think it's reasonable for someone to say that a grocery store shouldn't sell guns, or that a payment processor shouldn't process gun sales. Just for one example.

If you love guns, substitute something else you dislike. Currently people are aggressively pushing to prohibit the sale of adult entertainment, for example, and doing so by trying to get payment processors to block those transactions.

It's very common for people to hold the opinion that stores and payment processors should be opinionated as well and discriminate in terms of what they allow to be bought or sold.

And that's completely ignoring the question of how laws governing commerce interact with this. If you do business with a sanctioned entity the government's going to get angry real quick AFAIK, regardless of whether the business was itself harmless.

Freedom is valuable in games, especially text adventure games. And an LLM is a powerful way to give the player freedom, since it can respond to almost anything.

But for games to shine they also need carefully considered constraints - when the player bounces off constraints they start to understand how the game's world and systems function, which lets them build a mental model and be able to start thinking the same way the designer(s) thought and come up with solutions for puzzles or decide how to react to a given situation.

What makes a maze engaging as a challenge is that your path is closed off in some places and not in others. Ideally any maze also has one or more concrete solutions, so the player is rewarded for mastering the maze by finding an exit, or maybe finding objectives or creatures hidden at key locations in the maze.

You can probably use modern LLMs to construct this sort of world and set those constraints, but I wonder how far we are from being able to also have the LLM maintain that world state and enforce the constraints?

You're doing fine! In general the best trick for HN posts is just to add

extra

line breaks

so that there's enough space between your sentences.

I don't think the post you're replying to deserves what feels like an aggressive dressing-down. The reality is that post formatting on HN is really janky and it's hard to get it to do what you want. I don't blame someone for failing to master HN formatting on one of their first tries.

One problem with photorealism is a lot of players are on bad displays, or in bad viewing environments. Games often take this into account in their visual direction so that they will be more legible in these different environments. It used to be even worse when designing a game for say the Gameboy Advance or original Nintendo DS where you knew the screen wasn't backlit or wasn't particularly bright so your images needed to be bright and colorful. Even now, a Nintendo Switch game might be played on the bus.

For big budget games the solution for this is typically to have brightness calibration when the game first boots up, but the game itself still needs to be designed adaptively so that it's not Too Dark or Too Bright at critical points, otherwise the playability of the title is jeopardized. This runs counter to a goal of photorealism.

if they're working against the elected leadership of the US, who are they working FOR?

Unironically: Either themselves, or the American people.