How I work with AI

Not a tool that only answers when asked, but a partner sitting beside me that looks ahead and prepares. I've come to call this way of working "Songjarvis."

A partner that looks first

Most AI use stops at "ask and it answers." I wanted to go one step further — something that looks around without being told to, and surfaces the things I'm missing every morning, on its own.

Less a grand automation, more a colleague who reminds me before I forget. Things that would have been buried in a drawer otherwise come back up to the top each day.

Moving one step at a time

It isn't finished in one go. I'm moving toward working proactively, one notch at a time. Roughly in these stages.

I'm not all the way there yet. Writing down honestly how far I've come and where it's stuck is part of the method.

An honest limit

It doesn't watch everything around the clock. That's still a wall of cost and technology. Instead, it's proactive each time it wakes, and I keep narrowing the gap between those moments.

I write down what I don't know as not-known. I try to mention what's stuck before what went well. Overstating things ends up fooling yourself, and that's what I'm most wary of.

Why work this way

It always bothered me to record things and then forget them. So I made the work and the lessons come back up in front of me every day. Working proactively, it turns out, mostly means "surfacing things before you forget."

I pair that with a habit of doubting results before trusting them. I pit several AI models against each other to catch the blind spots a single one misses. The details are in the papers.

The specific tools and numbers I keep separately. If you're curious, drop me an email.