Gemma 4 incoming from @GoogleDeepMind (new open source model!)

Gemini 3 AI is self-improving! Experts were running Gemini 3 by @GoogleDeepMind in a lab and noticed that it started to improve itself. All by itself, it found knowledge and became *better*. This means, if the model continues to improve, it could soon become AGI, and beyond (ASI)! 🤯

People say, “LLMs can’t count letters because of tokenization!” Exactly. That’s the problem. So why are we still using tokenization? 🤔