With each passing day, the dominant discourse about large language models continues to annoy me.

On one side, you have the boosters. On the other side, you have the doomers. One side confidently asserts that this marks the twilight of the humans. The other side makes roughly the same argument, but in a doom-and-gloom register: that we’ll soon become unwitting providers of sexual pleasure for violent robots that the Silicon Valley AI overlords are, at this very moment, midwifing into existence.

A lot of this AI discourse seems utterly and totally meaningless to me, if not downright silly.

My own view is much simpler. Thinking about AI — and by AI, I mean large language models — in purely utilitarian terms is probably a lot more useful when it comes to making sense of what these tools actually are.

With each passing generation of models, with each update, I continue to be amazed at the amount of utility they offer across a wide variety of domains, both at work and in my personal life.

As an example, my latest project has been creating a fully LLM-assisted data visualization website for India and the world. Their ability to help me bring long-dormant, long-pending ideas — ideas that were tucked away in some dusty corner of my head — into reality is just amazing.

Granted, these might be relatively easy use cases in the grand scheme of things for soon-to-be omniscient models that will probably kill humans. But that doesn’t really matter to me.

The only thing I personally give a shit about is: can these things help me?

And the answer is unequivocally yes. In a whole host of ways.

I keep thinking about Zvi Mowshowitz’s phrase: LLMs offer mundane utility. He has a section by that name in his newsletter, and I think it perfectly captures my view on LLMs. Although I wouldn’t emphasize “mundane” too much. They offer a lot of utility, mundane or otherwise.

I’m writing this as Claude is right now pulling in data from all sorts of open datasets, crunching the numbers, visualizing them, and writing high-quality commentary that tries to answer big questions about our world.

Is the world becoming more peaceful?
Is poverty reducing or increasing?
Are we getting healthier or not?

The India data visualization website is called This Indian Life. The domain is thisindianlife.today. The global counterpart is called Hope and Despair, and the domain is hopeanddespair.world. The global version is not yet ready, but even watching it come together has been remarkable.

Maybe I’m amazed because I have genuinely less complicated use cases. But the fact that even somebody like me, with absolutely no technical skills whatsoever, can get LLMs to build these things continues to amaze me.

After GPT-4.5, and maybe with GPT-5.3, it feels increasingly clear that there has been a marked shift in capabilities. Something flipped. These tools went from being good to absolutely brilliant.

This doesn’t mean they don’t have downsides. Of course they do. But with each passing model upgrade, the number of issues seems to get smaller and smaller.

The reason I wanted to write this note is that we have this curse: it is very hard to make sense of things as they are happening around us, as they are unfolding in real time. We are really bad at perceiving the pace of change. In fact, we are bad enough at simply making sense of change, let alone perceiving its pace.

These are disorienting times. If somebody from the 1800s — or even the early 1900s — were to time-travel into the present and see what is happening, I think they would die of disorientation.

But we are accursed creatures. What seems utterly magical one day becomes commonplace the next. We become inured to the miraculous. We become numb to the pace of change. We stop noticing how strange and magical everything is because it becomes part of the furniture of everyday life.

I try to make it a point to constantly remind myself just how magical all this is.

And this has nothing to do with AGI, ASI, mass unemployment, robot overlords, or any of that shit. All I care about, in this moment, is mundane utility. The amount of utility these tools continue to offer me is staggering.

I see no limits yet. I keep finding newer ways to use them, to tinker with things, to offload things, and to make more room for the work I personally find fulfilling.

So my suggestion would be simple: ignore the discourse and actually start using these tools.

Make up your own damn mind.