A developer used coding agents to implement LLM inference on a reprogrammable FPGA, reporting that it simply worked. The post recounts a series of breakthroughs while entering hardware from scratch and argues that coding agents sharply lower the barrier to low-level hardware work, opening directions previously reserved for specialized teams.
FPGA
现场可编程门阵列
An integrated circuit that can be configured by programming after manufacturing, often used for prototyping or specific acceleration tasks.
Curated Intelligence
An engineering student open-sourced Specula, a dual-issue out-of-order RV32I processor implemented from scratch in hardware description language. The project documents the full journey from in-order to out-of-order design, covering register renaming, out-of-order issue and clock frequency optimization, making it a valuable open-source resource for learning CPU microarchitecture.
Altera announced Agilex 2X, an FPGA and SoC portfolio spanning high-performance, mid-range and power/cost-optimized devices, with Agilex 27 as the first family. Positioned for Digital AI infrastructure and Physical AI at the edge, it ships with a unified agentic AI design flow, though no process node, throughput or power figures were disclosed.
Altera has launched the Agilex 2X FPGA series, comprising four device families designed to support edge physical AI and digital AI infrastructure. The lineup features enhanced AI tensor DSPs, latest Arm cores, and high-bandwidth connectivity, with the flagship Agilex 27 expected to deliver engineering samples in late 2027.
OpenTPU is an open-source AI accelerator designed with the involvement of AI agents, exploring the limits of automated hardware design. Running on a Xilinx Kintex-7 FPGA platform, it successfully executes modern models like LFM2.5 and Qwen3.5, achieving bit-exact inference results matching its simulator. The project serves as a full-stack learning resource from high-level Python operations to low-level circuitry, demonstrating the feasibility of AI building its own inference chips.