Redwood

Redwood is the accelerator Architect Labs unveiled on 2026-08-27: from a high-level specification written by two human architects, an AI system generated its performance model, RTL, UVM verification environments, formal proofs, firmware, drivers and compute kernels, closed every block at 95% code and functional coverage in under two weeks, and took its first RTL drop to an AMD Versal FPGA with zero bugs (company site). It lives today on the FPGA; the silicon figures — a projected Redwood Nano on Samsung 8 nm serving Qwen3 at 49 tokens per second against a Jetson Orin Nano’s measured 28, at 1.3 W against 2.6 W — are projections calibrated from FPGA measurements. In this corpus it is the product of the first-of-authorship result and the substrate on which the loop reaches silicon: Qwen3 running on it found optimizations for its own operations.