2026 Life Update
Well, things have changed. I'm no longer doing Engineering Management at BitKernel, and now I'm back as a Machine Learning Engineer working on an AI Platform at Mozilla. This change happened in June of 2026, and I have just now had the opportunity to update my blog to reflect that change. I'm no longer employed as an Engineering Manager, and I'm back writing software. That beign said, the job has definitely changed a fair bit now that everyone is using AI to write software,
What does this mean for Baseweight
Baseweight in its current incarnation is in caretaker mode. The applications will eventually be unpublished, and the company will be fully wound down by the end of the year. I let videoscan.ai expire, but kept the baseweight.ai domain for two more years. Now that I'm working at Mozilla, I'm concentrating on fixing my health since that was something that definitely suffered while I was working on Baseweight. I definitely learned some painful lessons about startups, namely that incorporating is a good idea for tax purposes when you succeed, but if you don't even have a product and customers, it's just expensive cosplay that basically goes nowhere. I also learned that Vancouver is a terrible place to start a business unless you're in the outdoor, athleisure or real estate industry because you need proximity to your customers and Vancouver doens't really have too many industries. The people who are successful actually know people who have real problems that need to be solved and as a technologist, you're always stuck in the ineviable situation of having a solution to a problem that nobody has.
What next?
Well, I'm speaking at MozFest about OnDevice AI, which will mostly be the technical lessons learned from Baseweight, but also about the work we're doing at Mozilla in Firefox with features such as Link Preview. The talk will be mostly focused on the importance of running your own models, and how outsourcing your data and your thought process to large companies cedes power to them. This is funny, since I'm also very guilty of doing this. While I'll vibecode, I still won't generate writing using LLMs and I choose to still do this the old fasioned way by hand. I don't know if it's just about feeding the beast, or whether it's because I find it faster to just type what I'm thinking instead of wasting tokens on something that won't have the same tone and will be diplomatic and less mean.
I'll try to write more here soon.