I decided it was time for me to dive in and learn about AI for coding. It started at the beginning of Feb after I had taken an online cource fomr CEC and then watched the weekly Adafruit Deep Dive w/Scott (Friday at 5pm Eastern) and learned how he is using coding agents.
I thought I’d take on a somewhat simple embedded project and then things just exploded with goodness and I had to step back to figure out how I wanted to approach this.
Fun fact: The Raspberry Pi GPIO pins can only source about 16mA per pin.
Here are my overall guidelines:
- Since I want to test the edges so to speak I will try to push the agents to see what they can and can’t do.
- Do as much using the agent as possible – some people call this Vibe coding. I like neither the phase or the concept. That said, I don’t plan to employ this all the time for thing I can do myself but there are some thing beyond my skills that agents can do.
Before I get running I need to go over how I’ve named my agents.
Chelsea is ChatGPT
Tex refers to Codex
Jean Claude (JC) is Claude
Cici (CC) is Claude code
Right now I have a few projects on the go. I’ll give a brief description now and the rest will fill in over time.
- Set up a Seeed Studios Wio Terminal as a IR TV remote.
- Set up an Adafruit Funhouse as an IoT device that measures Temperature, Humidity and Barometric pressure. Push this data the cloud in a way that is useful and usable
- Build a system for the Meta Quest 2 that implements something like manim that displays and animates the output on the quest in 3D. This will require extending manim or perhaps the design of a new system that has 3D built in from the start.
One of the things I learned from Scott is the idea of using feedback loops so the agent can examine the results of it’s code. I started to look at hardware in the loop debugging and the possibilities and had a chat with chatGPT