The Early Days of Anthropic & How 21 of 22 VCs Rejected It | The Four Bottlenecks in AI | Anj Midha

By 20VC

Categories: VC, Startup

Summary

Culture is the most critical AI bottleneck—not algorithms or compute. Anj Midha, early Anthropic investor, reveals that scaling laws remain exponential in unexplored domains like materials science, and the real competitive advantage lies in building feedback loops that keep models learning in production environments.

Key Takeaways

1. The four bottlenecks in AI are: context feedback loops, compute, capital, and culture. Culture solves the algorithmic innovation problem by attracting mission-driven researchers who aren't locked into specific architectures.
2. Scaling laws haven't plateaued—they're accelerating in unexplored domains. Material science and superconductor discovery show 'super exponential gains' per iteration, while saturated domains like coding show diminishing returns.
3. The last-mile business advantage is in physical feedback loops. Deploy models in real-world domains, collect real-time performance data, and pipe verification data back into training—this creates defensible moats vs. pure pre-training.
4. Most frontier AI models underperform on physics and chemistry benchmarks. When Midha benchmarked Claude, Gemini, and others against applied physics problems at Stanford, major gaps emerged—indicating where competitive advantages exist.
5. 21 of 22 VCs rejected Anthropic in early days. Despite massive rejection, the company succeeded by focusing on alignment and safety culture—demonstrating that conviction in mission-driven hiring outweighs early investor consensus.

Topics

• AI Feedback Loops & RLHF
• Materials Science & LLMs
• Frontier Model Inference Security
• Scaling Laws & Compute Saturation
• Mission-Driven AI Culture

Transcript Excerpt

AI alignment, don't get me wrong, is hard, but not the hardest problem. Human alignment is really the problem right now. Our guest today is the most prominent AI investor in the ecosystem, an Midhar. Why is he the most prominent? Three reasons. Number one, he's one of the founding investors of Anthropic. Number two, he led AI investments for Andre Horus where he made investments in Black Forest Labs, Mistral, Sesame among others. And then third and finally today he's the founder of AMP where he ...