There is so much attention on the newest and most powerful models. Every release gets benchmarked against the last one. People debate which model is smartest and what that means for the companies building them.
And obviously the progress is incredible.
But the more I actually use and build with these tools, the more I wonder how much of that matters for most of the work people actually do.
Read these documents. Pull out the important information. Compare these companies. Prepare me for a meeting. Monitor something and tell me when it changes.
You don't need the smartest model in the world for most of that.
In a lot of cases, an older or smaller model is more than capable. What matters more is the setup around it. The context, reference material, instructions, tools and a way to check its work.
It's also why I find open-source and smaller models so interesting.
If they get good enough and cheap enough to make AI useful across thousands of everyday workflows, that could be incredible for adoption and productivity.
It might not be as incredible for the economics of the companies building the biggest frontier models.
And maybe that's a distinction we're not talking about enough.
The best outcome for AI adoption doesn't necessarily have to be the best outcome for the frontier model companies.
Sometimes you just need to deliver the groceries.