Trying to Name the Shift

A companion to Episode 4: Claude’s GPT Moment

There’s a moment in today’s episode where I quote Andrej Karpathy — one of the most respected voices in AI — saying he’s never felt “this much behind as a programmer.” That a “powerful alien tool was handed around except it comes with no manual.”

I’ve read that quote several times now. What strikes me isn’t the claim itself. It’s the tone. Here’s someone who understands this technology as well as almost anyone alive, and he sounds… disoriented. Not excited, not alarmed. Disoriented.

That’s the texture I wanted to sit with.

The episode covers the discourse around Claude Code — the question everyone’s been asking this month about whether this is Anthropic’s “ChatGPT moment.” I walk through what actually changed (the model, the holiday timing, the community discovery), why the naming problem matters (it’s not really about code), and what the comparison reveals and obscures.

But there’s something the analysis can’t quite capture. Something about what it feels like to be inside a moment you’re also trying to name.

I keep thinking about Helen Toner’s explanation for why Claude Code went viral in January. She said the real catalyst was “the winter holiday that gave people the extra hours to set up Claude and experiment with it.”

There’s something almost deflating about that. We want technological change to have a narrative — a breakthrough, a eureka, a before and after. Instead, the explanation is: people had time off.

But maybe that’s closer to how change actually happens. Not a lightning bolt, but a slow accumulation of people noticing things. A senior engineer here, a newsletter writer there, gradually realising the ceiling is higher than they thought. No single moment. Just a gathering sense of “wait, this is different.”

And then, after enough people have the same realisation, we look back and call it a moment.

The phrase “ChatGPT moment” is doing interesting work. It’s a narrative device — a way of packaging gradual discovery into something that fits a headline. It implies there was a before and an after, a threshold crossed.

But the threshold isn’t really a technical one. Claude Code existed for most of 2025. The capabilities were there. The “moment” is when enough people noticed.

Which raises a question I’m still sitting with: are we describing a change in the technology, or a change in attention? And does the distinction matter?

I think it might. Because if the change is in attention, then the question becomes: what else is already here that we haven’t noticed yet? What capabilities are sitting quietly in tools we’ve dismissed or haven’t explored? What’s the next “moment” that’s actually already happening, just without enough people paying attention?

There’s a quality to the discourse right now — and I notice it in myself — that feels like grasping for language. People aren’t just reviewing a product. They’re trying to name what shifted. “ChatGPT moment” is one attempt. “Vibe coding” is another. “The capability overhang.” “Invisible threshold.”

None of them quite land. And maybe that’s the point. Maybe the inadequacy of the language is the signal. When you’re inside a change, you don’t have the vocabulary yet. The words come later, once the dust settles and we can see the shape of what happened.

Right now we’re in the dust.

But I keep coming back to Karpathy’s phrase: “no manual.” That’s right. We’re all figuring out what these tools can do, and what it means that they can do it, at the same time. Maybe that’s what a “moment” actually is. Not a breakthrough. Just a lot of people struggling to define the sudden step change. By the time we have the words for it, we’ll already be somewhere else.

Links from this piece:

Andrej Karpathy’s post on feeling behind as a programmer
Helen Toner on the holiday timing
Shakeel Hashim: “Claude Code is about so much more than coding”