2 engineers + AI designed a working LLM chip in 2 weeks, no human in the loop
新智元 · wechat · 2026-09-01
ArchitectLabs' Redwood paper describes an AI accelerator designed, verified and deployed in just 2 weeks: 2 engineers wrote only a natural-language spec, and AI generated the full RTL, verification suite and firmware with zero human intervention. The chip runs open LLMs on an AMD Xilinx Versal FPGA, with estimated energy efficiency 3.4x Nvidia's Jetson; spec changes take just 48 hours to regenerate, and peak throughput hit 115 hardware changes a day with 0 bugs in first delivery. Notably, the AI running on Redwood v1 is already involved in designing its successor.
Separately, SemiAnalysis reports OpenAI engineers admitted they can no longer understand the assembly-level code AI generated for their in-house chip: OpenAI built a low-level kernel language, Gluon, atop Triton, where complex hardware instructions are fully AI-generated. The piece argues this dissolves Nvidia's CUDA moat — AI targets hardware directly — and turns hardware/software co-optimization into an opaque black box.
Related event: Two Engineers Plus AI Deliver a Working Chip in Two Weeks(2 posts)→
More from Infra
- China rolls out 'token factories' and a 2026-28 token economy action plan — nordicinst · 2026-10-16
- Vercel Sandbox Drives go GA: attach TBs of data in sub-second, fork instantly — cramforce · 2026-10-16
- NVIDIA launches RTX Spark for local AI on Windows PCs — ostrisai · 2026-10-16
- SkyRL ships Day-0 RL support for Beam with +62% LoRA decode throughput in vLLM — vllm_project · 2026-10-16
- DFlash2 delivers up to 3.03x decode-only speedup for Beam across five benchmarks — vllm_project · 2026-10-16
- vLLM's Beam serving kernels cut GB200 forward time from 8.15 to 6.47 ms at TP8 batch 128 — vllm_project · 2026-10-16