I'm Matt Wood, and this is For Your Information. A live list of riffs and links for you and your agent, drawn from what I'm reading, noticing, questioning, concluding, and revising.
An open-source AI accelerator implementation including RTL, ISA, simulator, compiler and profiler. Supports running Qwen3, LLaMA2.5 and Qwen3.5 models on Kintex-7 FPGA PCIe cards.
OpenAI and Synopsys Announce GPT-Synopsys for Chip DesignBoth address AI-specific chip/hardware design: openTPU provides open-source RTL and ISA for AI acceleration while GPT-Synopsys applies LLMs to chip design workflows — complementary approaches to AI hardware development
Real-Time Video Generation on TrainiumBoth address domain-specific AI inference hardware: openTPU targets FPGA-based acceleration while Trainium targets real-time video generation — both represent custom silicon/hardware approaches to AI workload acceleration
Jalapeño Shows Power of LLMs for Chip DesignJalapeño demonstrates LLMs used for chip design (RTL generation), directly relevant to the kind of hardware design pipeline openTPU represents — both sit at the intersection of AI and hardware design
Supports
Mini-AGI: Continual Learning on 8GB VRAMopenTPU running Qwen3/LLaMA on Kintex-7 FPGA cards complements the Mini-AGI goal of running capable AI on constrained hardware (8GB VRAM) — both democratize access to LLM inference on commodity/accessible hardware
Inside the Inference Hardware Revolution Of 2026openTPU is a concrete example of the inference hardware revolution described — an open-source FPGA-based accelerator capable of running frontier models like Qwen3 represents democratization of specialized AI inference silicon