I'm more bullish on Chinese LLM models that prioritize consumption, spending, and efficiency than on large, closed-source models that consume $25 per output. For me, the great future of AI isn't models that run at a loss, consuming VC funds to run, but quite the opposite: efficient models trained for specific tasks. It's amazing that today, with a $20 subscription, we can run models that require enormous data centers to operate, but at the end of the day, that doesn't scale and is burning through too much money. I wouldn't be surprised if this year we start seeing more cuts in those areas or if labs start making their models more efficient.