Developer uses AI coding agents to run an LLM on an FPGA chip
An independent developer describes using AI coding agents to program a field-programmable gate array (FPGA) to run a large language model, using AWS's F2 FPGA instances. The author, who had no prior FPGA experience, reports that the coding agents successfully generated the low-level C++ and Verilog code needed to configure and verify the hardware, producing working inference.
GoKawiil's interpretation of the reporting above, not reported fact.
The account suggests AI coding agents could lower the barrier to hardware-level programming, a domain that has traditionally required specialized expertise in circuit design and verification. If accurate, this could encourage more experimentation with custom chips for AI inference outside large semiconductor companies, though the claims come from a single developer's personal experiment rather than a formal benchmark or peer-reviewed study.
- A developer used AI coding agents to program an FPGA to run an LLM, despite having no prior FPGA experience.
- The experiment used AWS F2 instances, which provide cloud-based FPGA hardware access.
- The author claims the coding agents were highly effective at writing and verifying the low-level C++ and Verilog code required.
Xilinx Arty A7 FPGA Board — If tinkering with FPGAs for AI inference sounds appealing, the Arty A7 is a fantastic beginner-friendly board backed by Xilinx and widely supported by open-source toolchains. It's a great way to actually try your own vibe-coded hardware experiments instead of just reading about them.
See Xilinx Arty A7 FPGA Board on Amazon → Affiliate link — we may earn a commission on purchases, at no extra cost to you. Product picked by AI based on this article; it is not a tested recommendation.Source: ankitg.me — Ankit Gupta, 2026-11-20
Published there as: “Vibecoding an Inference Chip”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.