The AI Agent Moat Is Real, but Narrower Than You Think
Kevin examined AI agent investment opportunities and identified where moats actually exist. The sector's real defensibility lies not in engineering patterns—which open source reimplements in weeks—but in proprietary trajectory data from execution, integration depth with customer systems, and evaluation infrastructure. Companies like Harvey ($190M ARR), Sierra ($150M+ ARR), and Cursor ($2B ARR) compound advantages through data flywheels, while Meta's $2 billion Manus acquisition signaled that 147 trillion tokens of execution data across 80 million VM sessions justifies premium valuations where framework elegance and generic tooling offer no moat.