New (short!) blog post up: on OpenRouter's AI Model Rankings, I noticed a peculiar new LLM topping the rankings by a large margin: Hy3. I looked into the data and only became more confused. minimaxir.com/2026/05/open...
Who's Who of AI
Max Woolf
Senior Data Scientist at BuzzFeed in San Francisco // AI content generation ethics and R&D // plotter of pretty charts
https://minimaxir.com
What they're sharing
minimaxir.com
minimaxir.com
Articles & links
I coincidentially have a blog post on this minimaxir.com/2025/08/llm-...
Their own posts
Recent commentary
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.
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.
First time I've seen a disclaimer like this on a tool with obvious AI-generated copy.
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.
Asked GPT-5.5 to make a WEBP encoding workflow from scratch while keeping outputs perceptually similar to the source image and it did it in 10 minutes lol (15% smaller, 10x faster encoding)
asked GPT 5.5 to find all grammatical errors in my short blog post for tomorrow and it found over 100 of them