Also, "less tokens" is not always straight forward. I doubt it's a coincidence that the cavemen skill (or now proxy, I guess) has lots of numbers, but not a single benchmark on model performance or actual per-task token savings

For example one paper I remember found that without CoT, just stating your prompt twice increases model performance. With CoT, the same function is served by the CoT restating the important parts of your question. Something about which tokens can affect which other tokens in attention implementations

Once you get to a week of battery life like a Garmin or those cheap armbands without screen that becomes a much better feature

Though I think it fair to speculate that if he had taken the deal he would likely have more money today, at the cost of less status

Treating value as purely monetary profit is a perversion of that. To a certain degree it's inevitable as the number of shareholders grows, so seeing it in public companies isn't surprising. But even then, if a company focused on creating equipment for sustainable, pesticide free farming were to pivot into making Hellfire missiles that might be insanely profitable while still destroying a large part of what existing shareholders value about the company

AMD's software story is still a lot worse than Nvidia's. But patching up vllm to run one or two models you care about on AMD hardware is a much easier proposition than using them in most other fields of AI.

https://github.com/activeing123/mcptoon/blob/main/src/mcptoo...

Humans are very space efficient

The irony of selling an AI assistant with that premise is not lost on me

Asking the LLM for a review of the code would still have caught it

Publishing any out-of-copyright books would be awesome thing to do. But it seems none of the books in this order would fall under that, and it's unclear how many old books they are actually ingesting. Current reports are mostly talking about relatively recent books

It can't account for things that only come up during implementation. But you can explore much more of the design space in planning to reduce the opportunities for Claude to wander down the wrong path during implementation

I'm not sure what would replace it. A welfare state doesn't work if the state doesn't have the money. State-run economies don't have the best track record either