What if I can just do it faster than talking to you?
Sunday Letters
SubscribeSign in
What if I can just do it faster than talking to you?
Sam Schillace<br>Aug 02, 2026
Share
I was going to write a bit of a reaction to Noah Smith’s article on whether more intelligence will buy us much, but he covered the ground pretty well. The core ideas - that intelligence is spiky, that there can be things we don’t undestand but AI can, just because of how it’s organized, the differences in scale and tacit knowledge, all resonate with me. The one thing he didn’t touch on that I think is also underrecognized: quantity has a quality all its own, as they say. Even if there are domains where the problem is essentially “solved” down the point of irreducible chaos, sometimes having a lot more, or a lot faster, thinking can add new value.<br>We see this a little bit with code already. Not only are the models good coders, but they’re FAST. They aren’t always perfect, but they do build some fairly complex things with high reliability, and both the speed and reliability are growing quickly. Some human coders still want to review everything manually, but others are starting to use the models themselves to do multiple, redundant reviews, which they can do because they are fast and abundant (and don’t get bored).<br>When talking to people inside regular corporations, there’s a clear dichotomy even just of awareness of what the cutting-edge models can do. This makes sense, most people don’t follow technical developments closely in normal times, and these are much faster than normal times, so diffusion is particularly uneven.<br>But this gives rise to an interesting challenge. If I encounter someone who isn’t using the tools to do their work, and I can use them to get what I need more quickly than they can produce it, the path of least resistance is to route around them - AI hasn’t taken their job, per se, but it’s made the old way of doing it irrelevant. There are lots of examples of this happening through industrial history - jobs evolve as the technology does. It doesn’t pay to be a bottleneck when people can go around you easily.<br>My suspicion, and most folks won’t like this, sorry, is that not only will “code go first” as I’ve said before, but “everything is code”. We think of programs as a specialized way to interact with computers in a frozen, one-off, and expensive way: we have humans write software to do a specific task, like edit a document. But really, code is just … how you tell computers to do, well, anything. And models, being better and better at it, and very fast, are using code (and we are using code through them) to do all kinds of things. Things that didn’t make economic sense just a year ago, like writing a custom program for just a single business report, happen routinely now without our even knowing them. Code, and coding patterns will spread. I think it’s a huge mistake to say that models are good at coding but limited elsewhere - the umbrella of code-accessible tasks will only grow over time.<br>Putting this together, I think there are a lot of people out there happily ignoring these new capabilities, doing business as usual, not being particularly curious about the new tools. I think this is a big mistake.<br>You don’t have to believe that AI is conscious, or even especially “intelligent.” You don’t have to like the frontier companies or think they’re behaving responsibly. You can think the whole thing is wildly overhyped and that there’s a bubble somewhere in here. None of that changes the practical question: can these systems do useful work that matters to you?<br>Increasingly, the answer is yes. And once that happens, the important comparison isn’t between an AI and a person doing the same job. It’s between someone working the old way and someone who can summon a large amount of fast, cheap, increasingly capable intelligence whenever they need it.<br>That’s the part I think people are underestimating. The technology doesn’t have to replace you to change your job. It doesn’t even have to be better than you. It just has to make it easier for someone else to route around the parts of your job that used to require you.<br>I wouldn’t spend too much time arguing about where intelligence ends, or which tasks are intrinsically human, or whether this all eventually hits some hard ceiling. Maybe it does. In the meantime, the quantity is going up, the quality is going up, the cost is going down, and the surface area of things we can express as code is expanding.<br>That seems like a pretty bad moment to stop being curious.<br>Thanks for reading Sunday Letters! Subscribe for free to receive new posts and support my work.
Subscribe
Thanks for reading Sunday Letters! This post is public so feel free to share it.
Share
Share
Discussion about this post<br>CommentsRestacks
TopLatestDiscussions
No posts
Ready for more?
Subscribe
© 2026 Sam schillace · Privacy ∙ Terms ∙ Collection notice<br>Start your SubstackGet the app<br>Substack is the home for...