Last night, my mind wandered and I asked myself, “What’s the most interesting thing I want to learn?” My friend recently joined a robotics startup, which had me reminiscing about competing in FTC robotics in high school.
Since I now train language models for my day job, this naturally led me to impulse-buy an ROSpider after asking GPT 5.6 for a robot I could use to tinker with VLMs and VLAs.
GPT 5.6 actually recommended a much simpler robot with wheels, but this guy was by far the coolest-looking suggestion.
I also work with another type of spiderbot at Exa; this is not a coincidence because nothing is ever a coincidence.
I have very little idea what the SotA is, so I did the modern thing and asked my favorite frontier LLM to write me a personalized course. OpenAI already has the basic context on my background, so I just prompted it with my overall goal of learning how modern AI is used in robotics and gave it ARENA as an example of a high-quality online tutorial to emulate.
My first “capstone” goal is to get my robot to respond when I call its name (tbd), use its directional mic and camera to locate me, and then fetch a tennis ball I throw.
I figured it’d be fun to document my journey, both at the object level and at the meta level of seeing how well LLMs can teach relatively advanced technical subjects to an interested novice.