LLMs have no moat. Gemini catching up so fast proves that compute and data are the only real barriers.
Training starts with research ideas, but eventually hits RSI (Recursive Self-Improvement). Every frontier lab will hit an intelligence ceiling well beyond what humans actually need—at that point, they're all indistinguishable to us.
That’s why Anthropic is so defensive. Compared to Google, xAI, or OpenAI, they're compute and data-constrained.
Claude’s real edge isn't raw intelligence—it's the harness. Base model IQ will be commoditized soon.
Wall Street right now:
Memory stocks ($MU): "Okay, you’re printing cash and doing massive buybacks, but it’s not sustainable."
Mag 7 ($AMZN $GOOGL): "Okay, you’re spending billions on CapEx for a massive future moat, but where’s the ROI right now?"
These guys literally want their cake and eat it too.