We’re standing on the edge of something new.
I’ve been developing software since I was nine, and I sold my first product at nineteen. A little automation utility that swapped out a bunch of settings on your Mac whenever you changed wifi networks. SMTP servers, desktop backgrounds, screensavers, and a dozen other things.
WiLMa (that’s the ‘wireless location manager’ for those of you keeping score) wasn’t something that I wrote because I thought automation was the future. It’s something I wrote because I thought it’d be cool and it seemed potentially useful. And it was, for a few hundred users. It got reviews on websites, mentions on podcasts, and a couple of other companies ended up cloning it. Not bad for a few months of inexperienced work.
When I was in university, I took a couple of robotics classes. One was a ‘design from scratch’ kind of course, while the other was more of a ‘here’s a hardware platform, go write the code’ course. Both of them were interesting, but I never really saw a future in the field. Nothing ever seemed to work quite right, and at best it seemed like we were playing with toys. What was I going to do? Go write manufacturing robots?
These days, I run an app development company, and some of our biggest clients are large corporations that need automation of their internal IT processes. It saves a lot of people a lot of time, but fundamentally it’s just a long series of instructions. When Apple releases a new build of Mac OS, we need to do updates or everything breaks.
Automation, until now, has been useful in tightly constrained environments.
But we’re entering an era where machine learning models are a lot more powerful and a lot more interesting in what they can identify.
And when I put two and two together… I think we’re on the edge of something new.
I’m putting this online to document what I’m learning and what I’m building. This is all a work in progress. Maybe it falls apart, maybe I get bored of it tomorrow.
But I think we can step outside of the tight constraints and use recognition models and LLMs to help us create better control systems for robots and other types of automation. Much better control systems. The kind where we can step off the factory floor or the robotics lab and into the real world.
Like my buddy Gord says, I dream small.