Sequoia & Benchmark made $80M bet on “Specific Intelligence”
The “TikTok for Language acquisition” 2x ARR every 2 months
From General Agents to “Specific Intelligence”
Many of today’s AI startups are built around the idea of Agents—especially general-purpose ones. But a new wave of companies, backed by Benchmark, Sequoia, and Lux, is taking a very different path: customized agents, or what they call Specific Intelligence.
The philosophy is simple yet radical: Generalist agents are impressive but lack competitive advantage. Just as companies don’t hire “any smart person,” they need domain experts.
True differentiation comes from agents built for specific organizations—with specialized knowledge and proprietary data. As the company puts it:
We unlock your company’s latent knowledge to train custom models and deploy a team of internal AI agents that work directly for you. These agents are owned by you, deeply specialized, and continuously learning.
That company is Applied Compute, which today announced an $80 million funding round led by Benchmark, Sequoia, Lux, and Elad Gil.
While details weren’t disclosed, Benchmark’s Victor Lazarte reportedly led the $20M seed round at a $100M valuation in June. This new round suggests the valuation has surged to around $500 million, a 5x increase in just three months.
Applied Compute was founded by three former OpenAI researchers—all Stanford alumni.
Yash Patil (CEO) was a core member of the Codex team.
Rhythm Garg (CTO) contributed to o1, OpenAI’s reasoning model trained with RL.
Linden Li (Chief Architect) was a key contributor to OpenAI’s RL training infrastructure.
Applied Compute argues that being smart isn’t enough—progress comes from experts, whether human or machine. To gain an edge, agents need specialized knowledge, drawn from a company’s internal data. They call this Specific Intelligence.
Their approach: Applied Compute engineers work directly with client teams, embedding inside companies to build and deploy internal agent systems. The entire training stack, agent platform, and toolchain are developed in-house.
Two-thirds of the team are former founders with deep technical backgrounds—from top AI researchers to Math Olympiad medalists. Early customers already include Cognition, DoorDash, and Mercor.
“TikTok for Language acquisition” 2x ARR every 2 months
I’ve also been watching the education and learning space closely—and the market is massive. A few startups I’ve covered recently have been growing at an incredible pace.
