bandrami

Anything with a probability mass or distribution function has entropy; the state space of Markov chains just tends to get combinatorially big as the chain's memory increases.

There was for a brief period a Middle Path, called libiberty (now mostly subsumed into gnulib). It generated GNU-compatible shim functions for whatever system you happened to be on that translated those calls into the native calls. So you wrote one clean GNU-esque codebase and gnulib patched together essentially a translation layer for it when you built.

As an aside, back in the days of Stack Exchange I would always type out manually whatever answer I found to make sure I understood WTF I was adding to the system.

They moved the functionality from Firefox to Thunderbird IIRC, which seemed like a desperate attempt to get anyone to use Thunderbird.

Buzz was weird because its main impact was killing another Google product, Orkut

In the early aughts the web was largely anonymous (MySpace and Friendster were just starting out and Facebook was still a Ivy league only thing). We thought we wanted to be able to find and be found by specific people, but in retrospect I definitely regret that change. It was better when I was just a randomish string of characters.

IDK, I ran an RSS feed that had ads 20 years ago (it was a niche hobby blog and every fifth post was an ad for supplies for that hobby). It made enough to pay for the hosting.

I mean, go to an IBEW hall in a busy area and talk to the shop steward; he can probably start you tomorrow

The xAI Anthropic deal was pretty widely covered, but the press release is here: https://x.ai/news/anthropic-compute-partnership

The absurdly large discount was why Anthropic could claim a revenue-positive quarter for the first time. But that discount window ends IIRC in September. The problem remains that xAI announced the discounted revenue as "revenue before discounts".

That's $25B before discounts, and their discounts to Anthropic in the current quarter alone are $12B of that.

Because most of the capital assets depreciate over a period of 5 years or so so you have to make the expenditures back during that time or you can't book any revenue.

They have to find a way to make inference profitable and they haven't yet

Another silent inflation of token count. There's no force on earth that can overcome that incentive for the labs.

And the hyperscalers have made the (questionable) choice of depreciating that hardware over 6 years rather than the usual 4.

Yeah, it's definitely more of a consumable item than an asset, as the logistics guys would say.

Sort of; it's like the itemizer systems at airport security. They're entirely theater (in the sense of "for show") but that show is itself important in that it makes entire categories of attacks significantly less feasible.

Sort of like how simply having a uniformed police officer on a street corner reduces street crime significantly even if the officer never interacts with anyone.

I should expand that: the money goes from the end user to the lab and only in the cases where the marginal utility of the small app exceeds the cost.

IDK, get Springsteen to record an aid concert for them?

IDK I've got some rando Chinese product that looks pretty much exactly like a late-2000s iPod "classic" and has some fairly generic Android interface including a decent music player. I think it cost like $40.

Right but if we're going to a small-batch bespoke development model then "marketing" more or less disappears because why would I pay someone to talk to an LLM when I can also talk to an LLM? The money goes from the end user to the lab in this model.

> few users, no revenue or minimal revenue

The 900 pound gorilla has for a while been "where is all the output of these LLMs?", and this seems to be the answer; the number of apps on the various app stores has skyrocketed over the past two quarters but installs overall are actually down a little bit. So it seems like we've had an explosion of software that one or two people use. Which is a cool situation to think about but is pretty bearish for the "AI as a profitable business" idea.

> Its not my debt.

Neither was AIG's, until we all woke up one day and suddenly it was

> the companies do this because it works

Until it doesn't. I'm not only old enough to remember Enron, I'm old enough to remember the S&Ls.

Your house (hopefully) goes up in value. The datacenter goes down.

People said that in 2008, too!

Which is why lifecycle funds move you into bonds gradually as you approach retirement age

It's even less like attacking merchant vessels without a letter of marque than it is like theft, though

It's not really scale so much as circumvention; what got criminalized was circumventing protection mechanisms which AFAICT Anthropic didn't do

That's why copyright infringement is legally different from theft

They can't for the same reason that cab companies can't make their drivers employees: they would have to employ far, far fewer of them than they do on contingency.