In a traditional Printed Circuit Board Assembly (PCBA) workflow, you work with multiple disconnected vendors. This includes one for PCB design, another for fabrication, a separate component distributor, and finally an assembly house. Each step requires coordination, manual file transfers, part sourcing, and managing lead-time mismatches. Disconnectedness like this increases risk of delays and chance of errors. A fully integrated ecosystem, such as EasyEDA, streamlines this entire flow by linking part availability, footprints, PCB design, fabrication, and assembly in one system.

Machine learning inference does not always require large memory or a powerful CPU. When the model size is under a few hundred kilobytes, it is feasible to run inference on a general-purpose microcontroller (MCU). For example, in a recent project using a 24×32 pixel thermal sensor, we deployed a lightweight inference model to detect human presence. The model was able to operate at a reasonable speed on a 32-bit MCU using a compact framework.

Claude Code is the agentic AI software development tool I reach for when I want to accelerate my coding. When given the proper prompting and tooling, it is efficient at solving the problem presented to it. Through system prompts that demand test driven development, verification using browser automation platforms like Playwright MCP, and showing it where log files exist so it can troubleshoot on its own. This allows it to work for longer without bouts of human intervention in between. On green field projects, I have it focus on building a UI/UX prototype before I hook up a data backend. This lets it focus on UI without having to worry about the backend.