AI as a Matter of Course: Tunguz vs Alsop

In October, Stewart Alsop wrote that despite all the hype, he hasn’t found anything useful to do with AI. I read Stewart’s post, and I agreed with it. What do I use AI for anyway?

So I was very interested in Tomasz Tunguz’s post this week about how he uses AI more than 100 times per hour. He compared it to average daily smartphone use (144 times), and concluded that AI is probably more important than smartphones.

I decided to actually try using AI over the course of a workday, and tracked it. My use looks very different than Tomasz’s. But now I am left wondering if that’s because I’m behind the curve. Here’s my usage from 9am to 5pm yesterday, organized by type of usage and by category:

I created a metric called intensity to gauge the level of AI involvement in tasks. For instance, using Google Docs auto-formatting is a low-intensity action since it requires minimal AI interaction (just a single click). Conversely, setting up a locally running LLM on my machine involves extensive back-and-forth with AI and various other interactions, making it a high-intensity task.

At first I disagreed with Tomasz’s characterization of dictation as an AI feature. He might technically be hitting an API from an AI-based service for his dictation, but I have been dictating to my Mac and iPhone for years, since way before the introduction of generative AI. Then I re-read his post, and I caught this aside:

Two AIs process my voice : initially a dictation AI and a language AI that edits for brevity & clarity (both of these run on my laptop).

That is a very different type of speech-to-text than I’ve been using! So I spent some time trying to figure out if I could set up something similar on my own computer. The idea of having an AI listen to what I’m saying and then synthesize it into better, tighter narrative seems like a pretty powerful idea.

In about 45 minutes, with Claude's assistance, I downloaded and installed multiple Llama versions from ollama.ai, created a Python script for clipboard-based text editing in my terminal, and managed to switch between several locally stored models to observe the differences when using one over another. Then I used exactly that process to write this paragraph.

I screen recorded it, so if you're curious, you can watch the video here.

What stands out to me is that I use AI when it occurs to me, while Tomasz uses AI as a matter of course.

I often don't realize how to use AI in situations where it could actually be useful, since it's not integrated into my work habits. The real unlock of AI’s promised value will be when people make that mental step, and AI becomes a natural and default companion to a broad range of work tasks.

I suspect this is a muscle, like great delegation.

If you use AI daily, for work or not, I'd love to hear about it - not just occasional experiments, but habits that are now part of your default behavior. Hit reply and let me know!