Two things can be true simultaneously: a) Modern LLMs can count the amount of letters in a word despite the counterintuition of tokenization b) Google Search Overview's LLM can fail to the amount of letters in a word because it's a quantized LLM There's a nuance that tbh no one care about anymore.
It's both funny and logical that OpenRouter is now the place for running Chinese OSS LLMs without jumping through hoops.
AI companies really have got to stop only posting announcements on X. Just syndicate it to your company blog, it won’t hurt SEO dammit.
the fact that more people are defending the jqwik secret message to get AI to delete itself than are condemning it is concerning.
I kinda want to see what would happen (both technically and community-wise) if an agentic LLM ported WordPress from PHP to Rust.
The one objectively good thing about the Claude Fable 5 launch is that it implies OpenAI will release GPT 5.6 in a couple days.
First time I've seen a disclaimer like this on a tool with obvious AI-generated copy.
The new Siri AI features are clearly using some sort of search engine (Google) so wonder how strong Apple's privacy guarantees are.
A fun pricing nit with agentic LLMs on OpenRouter: in DeepSeek V4 Flash's case, the LLM provider is extremely relevant as DeepSeek is the cheapest by a large margin as their caching is extremely cheap. Unfortunately downstream agent clients may not let you choose a specific provider.
I accidentally discovered a new technique for agentic benchmaxxing that got another 1.5x-2x multiplicative speedup over code that was already hyperoptimized. This speedup has been consistent across projects of different domains without regressions. You are not ready for my OSS projects.
