Just when you thought hardware engineering was safe from the AI takeover, a demonstration has surfaced showing a supposed next-generation AI designing a complex drone flight controller from a simple prompt. The entire process, from a blank canvas to manufacturing-ready files, was reportedly completed in just 3 hours, 13 minutes, and 56 seconds. Because sleeping, eating, and coffee breaks are for carbon-based lifeforms.
The AI agent, given the speculative and frankly fan-fiction-esque name “GPT-6 Astra,” was shown operating directly inside KiCad, the popular open-source electronics design suite. According to the developer’s post, the AI handled the full workflow: creating the circuit schematic, laying out the printed circuit board (PCB), routing all the connections, and placing the components for an F405 six-layer flight controller—the electronic brain of a high-performance FPV drone.
What makes this more than just a fancy drawing exercise is the AI’s reported ability to self-correct. After its initial design, it autonomously ran Electrical Rule Checks (ERC) and Design Rule Checks (DRC) to find issues, then proceeded to modify the circuit and layout to fix them. The process allegedly consumed a jaw-dropping 52 million tokens. At the end, it spat out a complete set of Gerber, drill, Bill of Materials (BOM), and 3D model files, ready to be sent to a fabrication house. Of course, the design hasn’t been prototyped yet, so whether it can actually fly a drone or just spectacularly release the magic smoke remains to be seen.
This isn’t the first time we’ve seen AI tackle the physical world. As we’ve covered before, AI agents are already designing and checking 3D-printable robot parts from plain English commands This Free AI Agent Designs and Checks Robot Parts From Plain English . Moving from mechanical structures to the dense, unforgiving logic of electronics is a significant leap.
Why is this important?
The demonstration, if legitimate, represents a monumental step toward the full automation of hardware design. The process of creating a complex, multi-layer PCB is a highly skilled discipline that typically takes human engineers days or weeks. Condensing that into a few hours based on a high-level description could radically accelerate prototyping and innovation.
It suggests a future where describing a device is tantamount to creating it, lowering the barrier to entry for complex hardware development. For now, the role of the human engineer seems to be safe—someone still has to solder the thing and find out if the AI has a sense of humor about capacitor placement. But for how long is anyone’s guess.
