Dario needs mythos to be a moat

Labs close capability gaps with more compute, better post-training, and faster eval cycles
Open weight models like llama, qwen, and deepseek DO ship w/ capabilities on par with closed models a few months later, as free downloads
For you, that's a better model every quarter and no lab trying to finesse you into their max plans
The moat is decentralized compute where thousands of GPUs run as one supercomputer across Singapore, the US, Norway, etc
Dario's pitch has one benchmark cycle left