There are some strange ways to use AI, but there are also some uses that make perfect sense. In the latter case, it was only a matter of time before AI systems began to play a role in the design and construction of digital machines capable of running them. The team at Quilter AI is leading the way with AI software tools designed specifically for this task. AI is already helping us design CPUs, and now it's helping us design entire computers. The computer reportedly started on the first request.
Using an AI-assisted design workflow, the company built a fully functional computer consisting of 843 individual components. Known as Project Speedrun, this computer lived up to its name in every sense of the word. It took less than a week to go from concept to fully operational computer. The company says this work typically takes about three months.
This wasn't just a proof-of-concept computer either. It's a real Linux-based computer that spans two printed circuit boards and was designed by a single engineer using Quilter's AI platform. Computer design did not completely eliminate humans. Engineers continued to define the system schematics, requirements, and constraints. What AI did was take over repetitive tasks. By handling the “donkey work” of layout and wiring design, AI has made it possible to build real-world computers in less time than it takes traditional hardware projects to get to the first layout revision.
How AI helped design and build computers
Most AIs require human prompting before they can perform their tasks, and this project was no exception. Before AI was unleashed, human engineers defined the system schematic and selected all the components. Additionally, humans determined the electrical, physical, and manufacturing constraints that the design must meet. In other words, AI was not invented from scratch. Rather, they were responding to a clear scenario set by human prompting. These are more complex than your average ChatGPT prompt, but the underlying mechanics are still similar.
Once the AI completed the layout, the design was carried through traditional PCB manufacturing and board assembly processes. Engineers then performed a cleanup phase and made any necessary tweaks. Importantly, Quilter insists that “this was not a rescue operation.” The AI-generated core layout was left in place without the need for extensive redesign or rewiring. “Polishing” took only 38.5 hours. For comparison, the estimated time for a similar non-AI-assisted build is 428 hours, according to the company.
Once the design work was complete and the finishes were applied, Quilter's AI system helped create a computer that required only a fraction of the resources and time typically required for such a build. In this example, the end result was a computer running an NXP i.MX 8M Mini Quad processor, 2 GB RAM, 32 GB storage, audio, video output, Ethernet, and a PCIe slot.
Why a successful first boot wasn't the end of the story
Quilter says its AI-designed computer booted on the first try. That milestone alone is unusual in hardware development, but the project didn't stop there. While the first successful launch certainly proved the concept, the Quilter team behind the project wanted to show that it could do more than just launch. They wanted to demonstrate that this wasn't just a computer you could turn on, but a machine you could actually use.
After the initial launch, the system was configured to run a full Linux environment, allowing the team to observe how the computer behaved as the software began making continuous requests to it. This is one of the stages that can cause conflict among engineers. This is the point at which many early revision boards can fail or exhibit weaknesses. Typically, this is when subtle (or not-so-subtle) issues such as power delivery, signal integrity, or undesirable thermal characteristics all become apparent. According to Quilter, the AI-assisted board worked as expected at this stage.
As mentioned earlier, Quilter also wanted to demonstrate that this was a usable computer. During testing, I was able to successfully run YouTube, games (AngryBots), as well as productivity suites like Google Meet and LibreOffice. Not bad for the first computer designed by AI, which took less than a week from drawing board to first boot. Although all performance, testing, and development data comes from Quilter itself, the results still point to a future where AI participation in computer design is the norm.
