I tried using agents on a small Orange Pi and other small machines, and I’ve come to the conclusion that, unfortunately, it’s not feasible to do this in a practical way.
Of course, I started writing an agent that could run in such an environment (long timeouts, retries, etc.), but it’s a real pain.
For this to make sense, you need much more powerful hardware (a 5-year-old Mac mini is fine); the other issue is power consumption. Unfortunately, until you can run Mavericks 2 on your own hardware, it’s pretty expensive.
Paper notebooks certainly have their place. It’s the same debate as paper books vs e-readers. I personally don’t like how my handwriting looks and also don’t want to carry more things around.
My setup is WSL2 on a regular Windows machine — no GPU, so local inference would be painfully slow
Gemini 2.5 Flash free tier is genuinely good enough — 1,500 req/day, I use ~105. Quality is solid for content generation and analysis tasks
$0/month is hard to beat — I did accidentally rack up $127 when I used a billing-enabled API key (wrote a blog post about that lesson), but with free tier properly configured, it's been zero cost for months
If I needed more throughput or privacy-sensitive processing, I'd consider local models. But for my current scale, free tier Gemini handles everything.
I like computers that can run for several days or weeks on a single charge. Today, better processors and better screens could make this possible. Add solar panels and we would have weak computers that work like calculators. No power from the wall!
reply