PCB is brought to you by Fable 5
The author challenged Fable 5, an AI, to design a functional PCB with no manual intervention, aiming to see if a simple description could yield a physical product. While the resulting E-ink display board worked, the process highlighted AI's current limitations in complex routing and component selection. This sparked a Hacker News debate on AI's role in hardware design, weighing ease of creation against the eroding satisfaction of traditional craftsmanship.
The Lowdown
Driven by a desire to create a simple PCB without prior experience, the author undertook an experiment: to have Fable 5 (an advanced AI) design a fully functional board based solely on a plain English description, without any manual edits or verification. The goal was to trust the AI with both the design and the subsequent manufacturing costs, acting purely as an 'end customer.'
- Fable 5 quickly generated a design for an RP2350-based development board capable of driving a 1.54" E-ink display, complete with component selection.
- Initial AI-generated designs contained 65 Design Rule Checking (DRC) errors and incorrect component packages (e.g., wrong footprints for an SPI flash and a transistor), which Fable successfully corrected after being notified.
- The AI used Freerouting, an open-source tool, for automatic routing, which struggled and left many connections unfinished; Fable then hand-routed the remainder. Later, the author discovered KiCadRoutingTools, which routed the board perfectly and much faster.
- Ordering through JLCPCB was described as straightforward, with Claude assisting in preparing files and guiding GUI selections, despite some component substitutions. Five fully assembled boards cost 130 Euros (plus display costs).
- Upon arrival, the boards were functional, recognized by a laptop, and successfully ran proof-of-concept apps like a stopwatch and image album, leveraging the E-ink display's persistent images.
- The author expressed mixed feelings: thrilled by the ease of creation, but a sense of accomplishment was diminished due to bypassing the traditional learning struggle. They intend to continue 'YOLO-ing' hobby projects with AI while maintaining meticulous review for professional work.
- A sneak peek was provided for the next project: an NVIDIA Jetson Orin Nano-based tablet, also designed with Fable 5.1 and KiCADRoutingTools.
The experiment demonstrates AI's impressive capability to lower the barrier to entry for hardware design, enabling novices to produce functional electronics. However, it also provokes reflection on the personal value of the design process itself, and whether abstraction comes at the cost of genuine learning and satisfaction.
The Gossip
Algorithmic Accuracy & Apathy
Commenters generally agree that while AI can assist with simple PCB designs, it struggles significantly with complex, high-speed, or specialized electronics. They highlight issues like poor routing, incorrect component selection, and a lack of true "understanding" of electrical engineering principles, contrasting it with the human ability to reason about general intelligence. Despite these limitations, many acknowledge AI's potential for simpler tasks and its capacity to raise the 'floor' for hobbyists.
Effort vs. Ease: The Learning Lullaby
Many users empathize with the author's mixed feelings about AI-assisted design, questioning whether bypassing the learning process diminishes the sense of accomplishment and skill development. There's a debate about whether "YOLO-ing" hobby projects with AI is a positive development that democratizes creation or a step towards intellectual laziness that erodes the satisfaction of mastering a craft.
Practical Progress & Prospective Pathways
Despite acknowledged limitations, several commenters share their own successful experiences using AI for PCB design, particularly for simpler boards or specific tasks like generating Kicad files or extracting datasheet information. The discussion also touches on the 'tooling gap' and the potential future where AI could fully automate design, manufacturing, and even supply chain resilience, possibly through improved benchmarking and advanced agents interacting directly with fabrication services.