Deep Cogito has $43M in fresh backing. TQ Ventures led the Series A, joined by Benchmark, Nexus Venture Partners, Atreides Management, South Park Commons, and Zscaler, the cloud security firm taking part as both customer and strategic investor. The raise lifts total funding past $56M.
Founders Drishan Arora and Dhruv Malrana, ex-Google engineers who worked on AI Mode and AI Overviews, argue future intelligence gains will come from better post-training rather than ever-larger pre-training runs.
The lab’s research centers on Iterated Distillation and Amplification, which gives a model extra computation to reach stronger answers, distills the improvement back into its parameters, and repeats. Expensive reasoning gradually becomes internalized, so later responses cost less.
Pre-training supplies a model with vast knowledge and capability, Arora said, while post-training decides what that model can actually become. With frontier pre-training getting pricier, post-training efficiency is becoming a real competitive lever.
