An AI designed the wiring and code for a breadboard camera

Nov 14, 2026 Last reply: 3 hours ago 2 Replies
Jump to replies (2)

The starting question here is a good one: can a conversational AI actually design the wiring for an electronics project and write code that runs? Robert Feranec put that idea to the test by letting Grok handle both sides of a camera build, from pin connections to the final script.



On the hardware side, the setup pairs a Raspberry Pi Pico W with an Adafruit OV5640 camera breakout and a custom microSD card breakout, all linked on a solderless breadboard with jumper wires. A micro USB cable and an SD card reader round out the parts list. Grok supplied the pin mapping, and those suggestions were cross-checked against documentation from Adafruit and Raspberry Pi before anything got wired up. CircuitPython 9.2.4 went onto the board along with the OV5640 library bundle, the script was edited in Visual Studio Code, and terminal output was monitored in Tera Term.



The debugging loop is what stands out to me. Every runtime error went back into the chat, which returned revised code each time. Early runs hit a TypeError about chip select pin definitions, then a MemoryError from allocating a 262,144-byte buffer, then another TypeError over memoryview conversion. The fixes adjusted the pin definitions, shrank the buffer, lowered the image resolution, and switched to writing raw bytes to the card.



The outcome: the device captured photos and saved them to the microSD card, with files named img000.jpg and img001.jpg that checked out fine on a PC. The open question is whether AI help is genuinely viable for hobbyist circuit and software work, and the maker has asked the audience to weigh in on exactly that.



Project by Robert Feranec. Watch it on their YouTube channel:

formatting link


Cross-checking the AI pin suggestions against the official Adafruit and Raspberry Pi pinouts looks like the smartest step in the whole workflow. Trusting generated wiring blind would be risky, and verifying each connection means a wrong answer is far less likely to damage parts sitting on the breadboard. That extra checking time seems like cheap insurance for a setup assembled with jumper wires.

Since the MemoryError only cleared once the buffer size and image resolution were reduced, a next step could be streaming the image data to the card in smaller chunks rather than one large allocation. That approach might allow keeping the full sensor resolution while staying inside the RAM limit. Trying the same circuit with a different camera module that has a mature CircuitPython driver could also be worth exploring.

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required