A Los Angeles based startup has demonstrated just how far AI driven hardware design has come. Quilter says its AI designed a fully functional Linux computer with 843 components across two PCBs in just one week, and remarkably, the system booted successfully on its very first power up, as reported by Tom’s Hardware.
The project, dubbed Project Speedrun, centers on a custom Linux single board computer created with heavy assistance from Quilter’s in house AI system. According to the company, the entire design process required only 38.5 hours of human involvement, with the rest handled by AI over the course of a week. In a traditional workflow, a project of this complexity would typically take around three months and more than 400 hours of engineering time.
What makes the result especially striking is that the dual PCB system worked immediately. When powered on, the board booted Debian Linux on the first attempt, an outcome that even seasoned hardware engineers rarely expect. One engineer involved in the test reportedly reacted with disbelief when the system came to life without any debugging cycles.
The core advantage, according to Quilter, lies in how its AI removes one of the biggest bottlenecks in hardware development: PCB execution. Normally, engineers must carefully manage a repetitive three step process of setup, execution, and cleanup. The execution phase, which involves routing traces, placing components, and satisfying physical constraints, consumes the bulk of the time and is prone to human error.
With AI handling that execution work, engineers can focus on higher level decisions and refinement. Quilter says its system can even manage all three stages if needed, but the biggest gains come from letting humans guide the intent while AI does the mechanical heavy lifting.
Unlike large language models such as GPT style systems, Quilter’s AI was not trained on existing human designed circuit boards. Instead, it was trained by optimizing designs directly against the laws of physics. The company intentionally avoided human examples, arguing that real world PCB designs often contain inefficiencies and mistakes that could limit an AI trained to imitate them.
So far, that gamble appears to have paid off. The successful first boot suggests the AI produced a layout that met electrical, timing, and power constraints without the iterative trial and error common in human led designs.
Quilter’s CEO, Sergiy Nesterenko, says the long term goal is not just to match human engineers, but to surpass them. He argues that AI driven PCB design could eventually explore layouts and architectures humans would never consider, leading to better performance and efficiency.
Beyond speed, the company believes the implications are broader. By dramatically reducing the time, cost, and expertise needed to design complex boards, tools like this could lower the barrier to entry for new hardware startups. If Project Speedrun is not a one off success, it may signal a shift where designing computers becomes faster, cheaper, and far more accessible than ever before.
For now, the image of an AI designed Linux computer booting flawlessly on its first try stands as a compelling preview of what AI assisted engineering may soon make routine.