Phase 2 of the AI shift

I’ve been sitting at my desk for a few minutes contemplating what to write about this week. That’s unusual for me, usually I know what I want to write about when I sit down. It’s not for lack of subjects. Actually, it’s quite the opposite - there seems to be too much to write about, suddenly.

I have looked at the last two years of GenAI through the lens of the transition from desktop to cloud that started in earnest around 1993. It’s a good analogy in many ways - the shift now is as categorical and complex as that shift, will likely fundamentally change how we conduct the tech business as well as disrupt many non-tech business models, have unforeseen impacts, move in fits and starts, be underappreciated and overappreciated at the same time, and so on. I’ve written about a bunch of that.

There was an early phase of that transition that was exciting and fun - you could pay attention to almost everything that was happening in the industry in near real-time, you knew a lot of the folks working problems and could track their progress, and you could try most new solutions and keep up.

And then…that changed. I remember starting to feel like there was so much going on that it was harder (and then impossible) to really keep up with all of it. You could stay educated but had to start picking your battles, and opportunity cost started to dominate.

I think that’s where we are now with AI. It doesn’t mean all of the problems are solved or questions answered, but so many things are getting so easy to do now that there are solutions and new approaches everywhere that are hard to track. Just keeping up with base model development is hard, much less the ecosystem of apps around it (this feels like web serving tech stacks - eventually you had to make a bet and stay in that lane, it was just too hard to do all of it). Token costs are coming down, OSS is everywhere, fine tuning is cheap and approachable, app from text tools are getting good, and so on.

This is the beginning of a shift from the pure explore phase, to a phase that is a little more pragmatic (though still filled with a lot of disruption and exploration). There are good opportunities for pragmatic solutions - all the usual suspects like dev tools, discovery, media, productivity, but conceived through the lens of the new paradigm.

Don’t be overwhelmed. Try to stay current. Play with what you can. The rest of the world is slowly waking up to what we have been working on - even though it feels like press coverage has been endless, the reality is that most people are only now just starting to understand and use AI, just like the web before it. Soon, there will be an absolute torrent of usage, users, and new ideas and businesses.