“ An interesting conversation between Sebe, the founder of TS Circuit, and Matt, the founder of Atopile, discussing how code reshapes hardware design and how AI accelerates innovation. It is worth listening to the entire conversation, which has been reshaped in Chinese for easier learning.”
The following is a detailed briefing document of the podcast, reviewing the main themes and key points/facts, and quoting the original text:
Theme: This podcast discusses the future role of code in electronic product design, focusing on how a software-driven approach can accelerate hardware development and overcome current industry pain points.
Main Discussants:
- Sebe: Founder of TS Circuit (“React for electronics”), focused on making it easier for web developers to create electronic products using TypeScript and React.
- Matt: Founder of Atopile (a tool for designing circuit boards with code), former Tesla engineer (firmware for Mega Pack thermal control, testing suite), with a background in mechanical/mechatronics/robotics/computer science.
Topics and Views:
- Limitations and Pain Points of Existing Electronic Product Design Tools:
- “The gap between hardware teams and coding teams through PDFs and Wikis”: Matt, while working at Tesla as a firmware engineer, primarily understood hardware through system diagrams and schematics in PDF format, which were often stored in places like Confluence. “I was mainly doing code. I mean, I was primarily a software firmware engineer. So even when I wanted to pull up something like a system diagram or schematic to figure out what my firmware should do, I usually only had access to PDFs.” This workflow is inefficient and prone to errors.
- Altium License Limitations: The high cost of licenses for traditional ECAD tools like Altium creates information barriers between hardware and software teams. “A large part of this is due to Altium licenses.” This limits engineers’ direct access to and understanding of hardware design.
- Operating System Compatibility Issues: Most mechanical CAD and ECAD tools only support Windows, which is inconvenient for engineers using other operating systems like macOS. “Almost all mechanical CAD and ECAD only support Windows, and only a few can run on OSX, or you have to run parallels.” Matt’s personal experience transitioning from Windows to Mac confirms this.
- Closed Datasheets and Information Sharing: Many chip manufacturers require signing NDAs to access their chip datasheets, hindering innovation and information sharing. “You would think in electronics, even for open datasheets, like ‘Hey, I want to buy your chip.’ Obviously, they would give me information about it, but that’s not the case. If you try to buy BMS chips from some large manufacturers, you need to sign an NDA to get the datasheet to try using that chip and understand its functionality.” Matt expressed confusion over this, believing it does not help sell more products and is not meaningful in the face of reverse engineering.
2. The Future Vision of Code in Electronic Product Design (“Tony Stark” Style Development):
- Accelerating Innovation and Iteration: The core vision of the podcast is to achieve rapid design and iteration of electronic products through code, akin to Tony Stark in the movies. “Tony Stark in his workshop resonates with all of us, man. It’s not just you and me; I know it’s the same for Nion and Duncan… In the future, we should be able to design a car, a truly high-quality production car, in a matter of days.”
- “Explore More, Deeper”: Through code, engineers can explore design options more effectively and try more “outlandish” ideas that might be rejected by project managers under traditional processes. “Letting you explore more effectively so you can create better versions of products, or delve deeper into what your project manager would say you’re crazy for, that we can’t do a six-month R&D project, we have to deliver something.”
- Learning from the “Lean” Model of Software Development: Hardware development can learn from the lean methods of software development, reducing the high investment in physical prototypes and promoting earlier, faster experimentation. “In software, we have this very lean mindset; you can simulate a business in huge ways… I think that lean mindset in software is very productive. It allows people to dream. I think by bringing these tools into hardware, we won’t see more cars. Who cares about cars? We have plenty of them. It will allow people to dream like we do in software, but across the entire hardware field.”
- Openness and Collaboration: Code as a way to describe hardware will change the industry mindset, promoting open ecosystems and information sharing. “I think describing hardware in code will change this because it will change the industry mindset. When you have a vibrant ecosystem and shared values, suddenly it becomes profitable to get involved. You don’t want to be that manufacturer sitting on the laurels waiting for people to come to you while others are using something else.”
3. Deterministic Generative Design:
- Matt proposed the concept of “Deterministic Generative Design,” where given the same input, the code will produce the same output. This makes designs testable and verifiable. “I see it as deterministic generative design. You get generative design, and you get these boundary conditions, and it will randomly create something within that. I see code as a deterministic intermediate step for electronics. This means if you recompile with the same input, you will get the same output, which means suddenly it is testable, verifiable, all those types of things.”
- Reducing Engineer Burden, Improving Quality: This approach can alleviate the burden on engineers to create high-quality products, making the “most walked path” the standard for quality. “It greatly reduces the burden on engineers to create high-quality products, so you choose the most walked path.”
4. Standardization and Ecosystem:
- De Facto Standards: Certain chips (like Raspberry Pi Pico, ESP32, STM32, Nordic NRF52 series) are becoming de facto standards due to their ease of use and strong community support, especially in prototyping and low-volume products. “Now, the Raspberry Pi Pico has effectively become a de facto standard for us and some of our users, usable everywhere.”
- Impact of the Arduino Ecosystem: Arduino’s support for many chips (including ESP32 and others) has greatly simplified firmware development, becoming Matt’s go-to choice for non-production projects. “I usually say I’m a firmware engineer, but I don’t really like writing robust firmware. I like using Arduino. In fact, that’s one of the biggest deciding factors for me when designing any non-production new product; I need it to be simple and easy to use. So it’s the ESP32, the Raspberry Pi Pico I mentioned earlier, STM32, or Nordic NRF52 series, all now support Arduino. I’m unlikely to touch anything that doesn’t support it.”
- Platformio: Matt recommends using Platformio as an alternative to the Arduino toolkit because it is more powerful.
5. Applications of AI in Electronic Product Design:
- AI as a Code Generation Tool: AI (like ChatGPT, GitHub Copilot, Claude) can serve as a code generation tool, accelerating the design process. Sebe mentioned Duncan (from Jitex) found that Claude performed exceptionally well in circuit design. “Duncan from Jitex did a huge article on using AI for circuits, and he found that Claude basically did a fantastic job.”
- Limitations of AI: Matt believes ChatGPT still falls short in generating large, high-quality code but is very useful for small-scale tasks and as a drafting tool. “I usually find that once you get ChatGPT to do something larger, the quality isn’t as high as I expected, so I need to rewrite most of the structure, so it’s best for small-scale tasks.”
Key Ideas or Facts:
- The fundamental issue in electronic product design is that current tools and processes are based on PDFs and closed information (like Altium license restrictions and datasheet NDAs), leading to inefficiencies and innovation bottlenecks.
- Code is seen as the key to overcoming these limitations, enabling “Tony Stark” style rapid iteration and exploratory design.
- “Deterministic Generative Design” is a core concept emphasizing repeatable, verifiable hardware design through code.
- The importance of open ecosystems and standardization, with certain easy-to-use chips (like Raspberry Pi Pico and ESP32) and their community support becoming de facto standards, driving industry change.
- AI (especially Claude) shows great potential in generating electronic product code but still needs further development to handle more complex tasks.
Conclusion:
The podcast paints an exciting future where electronic product development becomes faster, more open, and more innovative by introducing principles and tools from software development (like code, open source, lean iteration, AI assistance) into hardware design. The challenge lies in breaking the constraints of existing tools and business models (like closed datasheets) to establish a more collaborative and transparent industry ecosystem.
Note: If you want to receive KiCad content updates first, please click the card below, follow, and set it as a star.
Common Collection Summary:
-
Learn KiCad with Dr. Peter
- KiCad 8 Exploration Collection
- KiCad Usage Experience Sharing
- KiCad Design Projects（Made with KiCad）
- Common Questions and Solutions
- KiCad Development Notes
- Plugin Applications
- Release Records