>Predicting things 5 years too early is often as useless as not predicting anything.
I dont think Zitron has predicted anything 5 years too early.
I think the people who saw the tech bubble and predicted in '99 got far too much shit for being right about what would happen and wrong about when.
They were essentially trying to put a timeline on when a mass delusion would collapse which is an inherently chaotic thing.
Meanwhile the people who thought that the bubble was just the new normal usually didn't get any shit for it because they were wrong in a "conventional" way.
Anthropic admitted last year to losing money on inference, it had negative margins. The margins have improved and are now positive but there’s still a significant cost. If plans aren’t being subsidized it would mean that the margin on inference is ~99%+ which would mean OpenAI and Anthropic should be wildly profitable but both are still losing money.
The most widely accepted estimates (though I disagree with them) are that Anthropic’s margin on API inference is ~70% from which people extrapolate what their token usage would cost via the API and compare that to what their plan costs.
re: increasing usage with resets, it’s because they’ve oversold usage and need to show that usage is growing. They’re increasing usage on fixed price plans without increasing the cost, the only plausible explanation is they have unused capacity. If they were capacity constrained then the last thing they would do is give away more usage for free.
Since almost everything eventually goes bust, I wouldn't agree that this was particularly valuable, unless the analysis proved out in other ways. It's like the saying that some economists have predicted 10 of the last 4 recessions. The analysis that leads to that prediction may or may not be valuable, but is only valuable if it predicts something, because otherwise how can you know it has any accuracy at all?
Author here. I'm a space engineer (PhD, 30+ patents; founded RKF Engineering and Kythera Space Solutions, sold both in 2025) — this analysis started as the technical appendix of a novel I'm writing. Happy to answer questions or defend the math. Full appendix with the complete derivations is a free download: https://dl.bookfunnel.com/zf9caalfyl
I got my C64 (actually C64C) when I was 7, switched to a PC five years later just as I mastered BASIC and was starting to learn assembly.
To this day I can recite from memory a rough memory map, say which POKEs to do to switch background/border/text colors to my favourite scheme (white on black), and name the addresses of a few basic KERNAL routines.
I have preferred vendoring for a very, very long time. Not only does it ensure that I can always build, it also ensures that I remain in control of the software I make and reduces support headaches when other people build it.
As an excellent side-effect, it also introduces a cost to including dependencies and encourages me to minimize them, thus increasing quality.
It's only theft when people pay Anthropic for inference in other to improve their own datasets. It's not theft when Anthropic grabbed basically all ebooks and web content on the internet to build their own dataset, without paying anything to anyone
I implemented a pretty simple iOS/macOS text editor using TextKit 2, and it was astonishing how many hacks, workarounds and manual implementations were needed to support even the most basic features.
That codebase is the polar opposite of a finely crafted pen, and the ugliness all comes from the API.
Is there someone out there seriously arguing that developing nations should just recycle and entirely skip modern landfills?
It's even worse than that. There are a bunch of nonprofits that are spending their limited funds building recycling infrastructure in developing countries where they could get a much greater ROI (in terms of environmental preservation and quality of life for the people who live there) by putting that money toward building traditional landfills.
No, it basically uses a fixed seed for the random number generator to generate tokens. The generated random sequence is just as random as it would be with a random seed.
I’ve been running Proxmox for quite some time in the home lab, and one thing I’ve learned is that some choices are easy to make on day one. However, they can get super painful to change once your home lab grows. I wrote up seven of the ones I think are worth planning early, including storage, networking, CPU compatibility, clustering, backups, and infrastructure dependencies.
Let's say Claude Code's system prompt is updated to recommend responding in Simplified Technical English. It'll work for the current generation of models. But Anthropic will train the next generation via RL on Claude Code traces as they do today. As long as they don't change their reward design, the next generation is going to be pulled towards Claudish again, because speaking Claudish gives higher rewards, so in the end the prompt doesn't really matter.
Ideally these things shouldn't work like this but this seems to be the sorry state of RL we're in.
> Stock buybacks artificially inflate equity value - cash rich companies buyback their stock just to deploy that cash and prop up their equity value.
There’s nothing artificial about it, it’s functionally equivalent to a qualified dividend that lets the shareholder decide when to pay taxes on it.
If there are fewer shares of a company and the valuation remains the same, the share price goes up. That’s 100% real. You may think it’s artificial and you’re entitled to that opinion, but money talks.
Back when writing 500 lines of code by simply typing it line by line from the "handover" booklet the C64 shipped with. Then running it and it magically made a frog-game appear where you hit "space" and the tongue of the frog needs to catch a fly. Then forgot to save on the 5,25'' disk and everything was gone. Good times :)
I dont think Zitron has predicted anything 5 years too early.
I think the people who saw the tech bubble and predicted in '99 got far too much shit for being right about what would happen and wrong about when.
They were essentially trying to put a timeline on when a mass delusion would collapse which is an inherently chaotic thing.
Meanwhile the people who thought that the bubble was just the new normal usually didn't get any shit for it because they were wrong in a "conventional" way.