No update this month

I’m skipping my normal retrospective this month, as I sold TinyPilot and am taking some time to figure out my next project.

Retrospectives will hopefully resume in a month or two!

My first two years as a bootstrapped founder went poorly. I could barely find any paying customers, and all of my businesses lost money. I began questioning my decision to quit my cushy Google job.

In mid-2020, yet another of my businesses had flopped, and it was only kind of COVID’s fault. Desperate for a distraction, I made a little contraption that controlled my home servers through my web browser. I called it TinyPilot.

Applications closed

Previous overview

I’m a blogger, and I often commission custom cartoons for my blog posts like this one:

The blog’s previous cartoonist was the awesome Loraine Yow, who worked with me for six years. She recently changed careers, so I’m looking for someone who can take over as the blog’s official cartoonist.

I saw that Meta released the Llama 3 AI model, and people seem excited about it, so I decided to give it a try.

I don’t have much experience running open-source AI models, and I didn’t see a lot of documentation about how to run them. I tinkered with it for a few hours and got Llama 3 working with Ollama, so I wanted to share my instructions.

Provisioning a cloud server with a GPU

To run this experiment, I provisioned the following server on Scaleway:

Highlights

• I worked with the TinyPilot team to lock down access to deployment secrets without interfering with our workflows.
• I learned from my mistakes to limit downtime when migrating services between platforms.
• I wrote my first compiler, albeit an extremely simple one.

Goal grades

At the start of each month, I declare what I’d like to accomplish. Here’s how I did against those goals: