Bloomberg just put the US-CHINA AI gap on a chart and yeah, it's getting obliterated:
> Kimi K3 is close to Fable
> ~70% cheaper per task
> Anthropic thought China was 6–12 months behind
> Chinese now has a cluster of fronter labs
> GLM-5.3 isn’t even included yet, which would favor China's side
US labs still have the frontier but China is making it impossible for them to charge frontier prices.
Quick, unscientific take: Kimi Moderato ($19) vs Claude Pro ($20).
Both in plan mode, Opus 4.8 feels like way more value than K3. One simple 2-line prompt blew K3's session limit without finishing the task, Opus did the same task more than once.
Although these equations belong to different branches of science, they describe remarkably similar ideas.
Clausius viewed entropy through the flow of heat, Boltzmann through the multiplicity of microscopic states, and Shannon through the uncertainty of information.
"Mathematics for Computer Science" is one of the best freely available textbooks on the subject. In more than 1,000 pages, it covers mathematical proofs, logic, sets, functions, graphs, number theory, counting, recurrences, probability, and randomised algorithms.
The book is freely available under a CC BY-SA 3.0 license. It was originally developed for MIT’s Mathematics for Computer Science course, and it is suitable both as a university textbook and as a reference for self-study.
https://t.co/Aga1yf0DIo