@ContextVM is one of the comfiest pieces of tech ever coming out of the #nostrverse. Typesafe, uncensorable, global, instantly monetizeble backends for everyone. Extremely fun to develop too. I'm looking forward to see a real market for resources built on this. There is lots of cool stuff coming down the road I'm sure. View quoted note →

The great European self suck economy. Incestuous, masturbatory, meaningless. The collapse can't come soon enough. View quoted note →

GM

GM Can we stop beating around the bush and admit that code is not only written less and less but also READ less and less by developers? The best way to be token efficient is not going through the hoops of changing the input media but either using/having a language that is token efficient by nature (we're so back brainfuck chads) or by straight up programming (or letting program) with embeddings and vectors.

Can you save on LLM tokens using images instead of text?

What happens if you convert your prompts to an image before running?

I never want to hear Americans bitch about Deepseek "Tianamen Square" prompts on HN again. View quoted note →