A DeepMind researcher concluded that the frontier is no longer something money can buy, because a 300-person lab compressed frontier training out of scarcity — cracking the compute-moat thesis behind $650 billion of investment.
The skeptics answered with falsifiable calm: K3 still loses to a five-month-old Mythos Preview, Anthropic sits on a 10-trillion-parameter model, and the true frontier stays legally sandbagged.