The Chart That Launched a Thousand Pedants

Someone posted a graph this week. Model version number on the Y-axis, release date on the X-axis. The line goes up and to the right. The title called it “not quite exponential, but progress is progress.” It was a shitpost. A pretty good one.

The comments, predictably, split into three camps.

First, the people who got it immediately and just typed “lol.” Second, the people who genuinely started analysing the graph before realising the Y-axis was just sequential model numbers. Honest mistake, to be fair. I probably would have done the same. Third, and most entertainingly, the people who did not get it and then got very annoyed at the first group for saying they didn’t get it.

That last one is its own thing. There is a specific kind of comment that goes: you think you’re so smart, but there are a thousand subjects where YOU wouldn’t get it, followed by some declaration about everyone upvoting the original person being an ass. It’s technically a fair point. It’s also a completely disproportionate response to a meme about AI model numbers. The reply to this was “it’s a meme, calm down bro,” which is the correct response.

The GTA 6 tangent was funnier than it had any right to be. Someone asked whether Claude could just build GTA 6, which led to someone else suggesting it might trigger safety guardrails, which led to someone else actually defending how developers do use AI: scaffolding, debugging, documentation, asset generation. All accurate. Solid explanation, dropped into a conversation that had already moved on to jokes about Rockstar’s release schedule. The internet in microcosm.

There’s something genuinely interesting underneath the noise, though. The joke works because the model numbers are confusing. Claude 3 Opus, then 3.5 Sonnet, then 3.7, then 4, now Claude Sonnet 4, Claude Opus 4. Anthropic’s versioning makes a kind of internal sense but communicates almost nothing to anyone who doesn’t follow this stuff closely. Someone in the thread pointed out that the gap between 3.5 and 3.7 on the meme chart had the same spacing as the gap between 3.7 and 4.0, which is: yes, correct, that’s because it’s a joke graph.

I’ve been using Claude fairly regularly for work. The improvements between versions are real and they’re not subtle. That part isn’t a meme. Whether the pace constitutes “exponential” progress depends heavily on what you’re measuring and what your priors are, and anyone who gives you a confident answer either way is probably selling something.

The person who made the graph admitted it took several attempts with one AI tool before switching to another to get the result they wanted. Then someone else said they should have just used Excel, which would have taken ten seconds. Which is true. But also: the iterative prompting, the tool-switching, the “good enough” output at the end – that’s just how most people actually use these things day to day. Messily. Pragmatically. Not waiting for the perfect tool or the perfect prompt.

That’s probably the more honest story of where AI is at right now. Not the exponential curve. Not the imminent singularity. Just: occasionally useful, sometimes frustrating, weirdly good at some things and bafflingly bad at others, and almost always sparking an argument in the comments about whether the Y-axis is properly scaled.