Top Tweets for #ometv(()())()()(***)

One more thing people keep misunderstanding: CetinLM being trained mostly in Turkish today does not mean it is meant to stay Turkish forever. The earliest experiments actually tried a much broader multilingual direction. I eventually narrowed that down to Turkish and English for a very simple reason: you should understand the language you are training. Training on a language you cannot properly read, judge, or culturally interpret means you may not even notice when the model is learning badly, hallucinating, collapsing stylistically, or absorbing broken data. That is not serious engineering. Turkish is my native language. English is… let’s call it my working language. I can read technical material, follow documentation, understand most written content and evaluate model behavior reasonably well — but I’m definitely not claiming some elite, native-level English fluency. My spoken English is much weaker than my technical English, which is exactly why I’m careful about how far I push language coverage without proper evaluation. And the actual Base-v1 corpus reflects that decision pretty brutally: 🇹🇷 Turkish: ~93.74% 🇬🇧 English: ~6.26% Yes. Only about 6% English. English was never there to dominate CetinLM. It was there as a controlled secondary language — enough to widen the model’s linguistic space without pretending I could properly supervise dozens of languages I don’t understand. So if CetinLM’s English turns out surprisingly good after all this… I’ll probably be just as curious as everyone else. The long-term direction can absolutely become multilingual. But I do not want the lazy version of multilingual AI where you dump translations into the corpus and call it global. If CetinLM expands into more languages, I want that expansion to be engineered properly: language quality, cultural context, evaluation, behavior, and real usability. The goal is ultimately to make the system usable by people everywhere. That takes time. So if you’re already imagining yourself using CetinLM in your language, you have two options for now: be patient… or start learning Turkish. First, let the product exist. Then we make the world bigger. 🔥 #CetinLM #AI