My understanding is that most testing methodologies can have false positives, it's just that some methodologies are extremely unlikely (like 1 in 100,000,000). If you apply such a test to the 10 people who were at the scene and get a match, you can be pretty confident that you have a legit match. If you apply it to all 300,000,000+ Americans and get a match... well, the only thing surprising about that would be if you only got 1 match!

8/19/2025:

SAM ALTMAN: There’s a lot of things we could do that would grow faster, that would get more time in ChatGPT that we don’t do because we know that our long term incentive is to stay as aligned with our users as possible. But there’s a lot of short term stuff we could do that would really juice growth or revenue or whatever and be very misaligned with that long term goal. I’m proud of the company and how little we get distracted by that, but sometimes we do get tempted.

CLEO ABRAM: Are there specific examples that come to mind? Any decisions that you’ve made?

SAM ALTMAN: Well, we haven’t put a sexbot avatar in ChatGPT yet.

'X is the #1 news app in the world' according to people who get their investing advice by listening to the ranting homeless guy on the corner.

You are absolutely correct about it being better to use a CAD tool when making dimensional, mechanical parts... although perhaps not to the degree that you might think. If you're already familiar with Blender, a complex Modifier stack or even a set of Geometry Nodes will give you the perfect meshes, precision, and flexibility that you would expect from a CAD tool. If you're getting started and need to do that kind of work, definitely use a CAD tool. Blender is an artist's tool and the learning curve to be able to use it with that kind of mechanical precision is steep, but the functionality is there.

He's never really been a democrat or a republican, just a wannabe autocrat who aligns with whatever group gets him closest to power.

As long as we do away with all of the tax loopholes that allow the rich and powerful to avoid paying their taxes, I'd be perfectly happy with your suggested UBI + Mostly Flat Tax suggestion.

Yeah, and it's worth emphasizing that that's $36k over 3 years. The median rent in Texas is about $2k/month and in Louisiana it's around $1.5k/month. That $1k/month is certainly helpful for people, but it's not like it's going to free up many primary wage-earners to go back to school or to start their own businesses or make some other investment in their long-term success.

Do you want a Reticulum? Because this is how you get a Reticulum.

What percentage of video game aren't covered by the description of Bad Guy messes up the Setting, you have to fix it and fight Bad Guy? The Zelda games are especially repetitive, but that basic formula is everywhere.

It looks to me like they did get the message: "The cuts mostly effect managers and contractors."

I generally like this idea, as long as it's written in such a way that the content creator must be a currently living human who significantly contributes to the creation of the piece. Any idea about how to handle the use case where a work is created by multiple people, such as co-authors? My first thought is that one of them would need to be elected as the copyright holder, but that creates a corner-case problem if that designated copyright holder dies prematurely, because now the surviving partner will lose their copyright protection.

I'm with you that containers are sweet, but we're still a long way away from a world without server virtualization. Virtualization took off because it could do just about everything that you were doing with physical servers, but it could do it cheaper while adding significant functionality (especially around availability). Through p2v, there was a relatively direct path into that virtualized space that required a few hours of labor per server being moved, but no significant architectural rework. Containerization asks a lot more of the customer and pretty much requires in-house developers (which means that it's most practical for in-house solutions). I haven't seen a lot of movement from major software vendors towards making their tools available as Containers; instead I've seen a *big* push towards SaaS, which offers the vendors the ever tempting subscription revenue that started this whole discussion.

If we are in a simulation, who is to say that this simulation accurately reflects the real world that's running it? If you were an NPC in Minecraft and your perceptions and life experience was entirely contained within the game, you would think that that was perfect reality.

The way I see it, the natural result of a simulation theory is a cascading sequence of less real realities where the residents of each think that theirs is real and eventually create simpler, abstracted realities for their games... which eventually hit a sufficient level of complexity to generate their own even simpler, even more abstracted realities. None of those child realities are perfect (despite the perspectives of their residents) though, and so you end up with a sequence where we can predict that the simplest reality is one that cannot yet run child realities, but we don't have any meaningful way to speculate about what the other end of that spectrum might look like.

It's worth pointing out that Elad's "Less to More Regulation" label could be accurately changed to "Less to More US Production Expenses" without changing anything else on the graph. There are *at least* two explanations for the trend that's been observed. This article strikes me as searching for data to justify an established view-point, rather than arriving at a conclusion based on the available data.

The answer to the question they imply is pretty clear on this one: the machine is flawed, but responsibility ultimately rests with the human operator whose entire job was to weed out that sort of flawed content. It doesn't matter if there was one article or a handful, any way about it, the human failed to manage the AI properly.

More interesting than "who's at fault" is the question of "why did the fault happen?" I suspect it's one of two situations. Either the human operator got lazy and stopped doing their job, possibly because the AI was so good that they grew complacent... or the human operator was completely overwhelmed by an incredible volume of AI generated BS and this stuff slipped through the cracks as they were busy eliminating the truly bizarre stuff. Either way, it says something interesting about the state of these AIs!