Applied Compute helps Fortune 500 companies build, deploy, and continuously improve AI agent workforces trained on their own proprietary data and workflows — what the company calls "Specific Intelligence." Founded in May 2025 by three ex-OpenAI researchers (Yash Patil, Rhythm Garg, Linden Li), the company embeds its own engineers directly inside enterprise environments to design training setups, post-train open-weight models using reinforcement learning, and ship autonomous agents into production. The key differentiator is a continual learning loop: every agent decision feeds back into training, so agents get better the more they're used — all within the customer's own secure environment. The core platform (AC2, launched August 2026) lets customers post-train models against their existing production harness. Buyers are F500 enterprises in sectors where proprietary workflow knowledge is a competitive moat — think financial services, healthcare, and complex operations. Revenue appears to be a mix of high-touch forward-deployed services and a recurring platform subscription.