Yeah, I'm amazed at how quickly I spotted this one. I didn't even read any prose since the first section header was visible before I started reading and somehow my brain jumped there first and it was a dead giveaway.

I think I may be developing defensive scanning strategies, that are probably useful for avoiding slop, but not good for my focus or general reading comprehension.

My first run got Korean Peninsula in 3064 BCE...

That is surprisingly similar given how few people seem to have commented so far. Makes me wonder if some part of this is starting from a constant seed or something...

"Any Korean Ever"

"You have died of Dysentery [on the Korean Peninsula in 3064 BCE at 2 months old]"

I was actually a bit surprised since it also said I was exclusively breastfed which seems to be fairly protective against diarrhea. I have a feeling this didn't account for all the joint probability distributions...

All babies were exclusively breastfed in 3064 BCE.

Yeah, it doesn't give totally absurd combinations (like a male dying of birth complications at age 10), but the probabilities are still likely rather rough approximations (eg. sorting cause-of-death distributions to a few bins like "infant", "child", "adult", "elder").

Strangely, I too, died in the Korean Peninsula at 8 months old, but circa 6000 BCE.

Hmm, I was at the Korean Penisula too. How representatively random is this thing?

I died at age 36 of a workplace accident in the Yangtze delta in the year 700

I died of malaria at age 1 in the West African Sahel!

Babies died of diarrhea anyway. Breastfeeding is not a panacea. If it were, doctors would simply prescribe a flask of breast milk, and mothers would earn money by prolonging lactation for Big Medicine.

I died at 1 in Mesopotamia in 3966 BCE of dysentery.

I mean, they were already capacity constrained and just introduced a larger model that takes more capacity to run... They were gonna have to hand some business to competitor one way or another.

> I mean, they were already capacity constrained and just introduced a larger model that takes more capacity to run

I am willing to bet that the SpaceX deal is probably why Fable's launching now, as they are much less compute constrained than they were a month ago.

Is it a larger model or just better trained? Anthropic does not actually claim it is a larger model anywhere that I can see.

Price is based on perceived value, not cost to produce. There is no international court of price justifications; if customers are willing to pay $X you can charge $X.

That and a model can be the same size, yet use a lot more compute, I guess think of it as intelligence per watt used or something like that.

Exactly. The company should care because it drives margins. But pricing to customers should not change unless it was artificially high (competitors offer more value for same money) due to profitability concerns.

Opus 4.7 was smaller and people still paid 4.6 prices.

gpt-5.5 isn't larger than gpt-5.4 but costs double.

Yeah, I do the same, but without the catchall for exactly that reason. If I start getting spam, the e-mail gets disabled.

Smart, what server / service do you use?

Look up email alias service or something similar, if you aren't looking to self host. I can't recommend the service I use, because I'm grandfathered in to my plan, and their current plans for new customers suck, but there's enough providers out there that you should find something competitive.

If you want to 'self host' on a provider, I thing cheap/free options are available from cloudflare, Google, and similar enterprise companies.

If you want to truly self host, I don't have experience, but this guy who does gave a great thorough answer for those who are interested: https://news.ycombinator.com/item?id=48073510

I'm paying $15/yr currently for a catchall, plus the domain. I think new customers get charged $50+ a year, maybe even closer to $100.

A recent NBC News poll gave "AI" a -20% net approval rating. If you're in the U.S., the people you run into IRL are kinda weird.

(I didn't quickly find polls for the rest of the world)

Perhaps they are, but I think what I think is odd is how many are simultaneously opposed to the effects of AI (data centers, environmental impact, job loss) yet also are daily, paying users of it even outside of work. I know several who are like that.

I wonder how many in that poll have a negative sentiment about AI, yet are also heavy users. There seems to be some kind of disconnect going on?

> I wonder how many in that poll have a negative sentiment about AI, yet are also heavy users. There seems to be some kind of disconnect going on?

Being a heavy user seems like it'd create a lot of resentment if you don't actually enjoy doing it.

If you have used a tool for years and years and suddenly it shoves in a bunch of AI, like Duolingo or Google, are you a heavy user who may well dislike the results?

I have to put in some work to not be a heavy user by any reasonable definition.

ChatGPT has been downloaded >1 billion times on the Android PlayStore. AI is incredibly popular.

People have all sorts of concerns about how AI will change society, but that's precisely because it's so useful for so many things. If it were useless or just a fad, there would be no reason to worry.

The startup where I worked had 1 million downloads (which is much less), but how many of those were using it daily vs trying it for 30 minutes and forgetting about it?

Less than 10% in our case :)

1 billion downloads, even if only 10% of those are active users, is a massive number.

Heck, I haven't even downloaded the ChatGPT app, but I use their website all the time.

AI use has taken off, and a large fraction of the population is using it regularly, if their own accord.

Yep, the free tier. Will they pay for it or stick around once it becomes more ads than anything else?

My point was that using something heavily means you have an informed opinion about hating it. Hating something you have never tried is way weirder.

Maybe we use search differently, but I very often don't just want an answer, I want to find a website to help me. Maybe it is because I need to do business with a company and need to find their website to interact with them, or maybe I saw a cool site awhile ago that's relevant to what I'm doing now and didn't bookmark it (because I dropped that habit when Google search was good), or want to read the official documentation about a product I bought, that someone already put a lot of effort into making complete enough and digestible to a wide audience... and the LLM responses tend to get in the way.

Like the parent I use good/paid AI when I want an AI response. So, yeah, an omnibox that knows when I want "an answer" and one that knows when I want to find a thing sounds slightly more convenient than switching between two tools, but Google search is not that Omnibox.

Yes, but more specifically, it's 20% on top of your marginal rate on your capital income which maxes out at 20% federal in the U.S. for long term gains However, it is much closer to 0% for the most wealthy Americans because they never realize their gains, which is the only time the U.S. taxes capital gains. They just fund their lifestyles with debt against their assets. Then when they die, their heirs get a basis step up at death.

Graham gets this totally wrong, adding the 20% to 37%+4.75%, which are rates applicable to labor income (and short term capital gains, but those are very rare among the most wealthy Americans). That is such a major error it is hard to take any of the argument seriously.

Edit: Updated account for short term gains.