Claude Code is about so much more than coding
It’s a general-purpose AI agent. And it’s already a pretty good knowledge worker
Over the holiday break, the news spread like wildfire. The presents under the tree were ignored — instead, the AI world collectively sat down to give Claude Code a try. And, almost without fail, they were blown away.
“I’ve never felt this much behind as a programmer … Clearly some powerful alien tool was handed around except it comes with no manual,” said AI legend Andrej Karpathy. “I gave Claude Code a description of [a] problem, it generated what we built last year in an hour,” said senior Google engineer Jaana Dogan. Everywhere you looked, the vibes were the same: Claude Code with Opus 4.5 is a game-changer.
It is all too easy for non-coders — policymakers, journalists, and knowledge workers in general — to hear this and shrug. “AI can code” sounds like a development that affects software engineers, not the rest of us. When your job doesn’t involve GitHub repositories or terminal windows, a better coding assistant seems like someone else’s problem.
This misses something crucial. When people hear “coding,” they picture a specialist skill for a specialist domain. They don’t picture booking theater tickets, analyzing spreadsheets, or processing invoices. But those are all tasks that happen through software — which means they’re all tasks that can be accomplished by an AI that can write and execute code. They are, in fact, all tasks that Claude Code has done for me in the last two weeks.
Code, after all, is just a language by which we instruct computers to do things. An AI agent that can code, then, can … do almost anything you do on a computer. The question isn’t “is this a coding task?” It’s “can this be done digitally?” If the answer is yes, there is a very good chance that Claude Code can do it.
This is crucial to understanding why Claude Code has implications for everyone, not just the developers that have already been wowed by it. Claude Code doesn’t just generate code for engineers to review and deploy. It uses code to accomplish tasks. The “Code” in its name is misleading, and undersells the actual product: a general-purpose AI agent that can do almost anything on your computer.
Combined with the “Claude in Chrome” extension, which lets Claude Code control your browser, the result is something like having a very smart generalist at your command, 24/7 — one which never gets tired, never gets bored, and can do almost anything you ask in a fraction of the time it would take a human.
I have absolutely zero coding experience. But in the past two weeks, I’ve had Claude Code go through my bank statements and invoices to prepare a first draft of my tax filing. (It got everything right.) I asked it to book me theater tickets: it reviewed my calendar, browsed the theater’s website for ticket availability, and picked a date that had good availability and suited my schedule. It built me a series of automation tools that will collectively save the Transformer team about half a day of work each week. It planned a detailed itinerary for a forthcoming vacation, including extracting hundreds of restaurant recommendations from my favorite influencer’s Instagram highlights.
I’ve had it reformat spreadsheets, produce detailed research reports, and help me write articles (including this one). It’s built me a beautiful new personal website, a custom dashboard for browsing Audible, and a script that syncs Letterboxd to my personal notes app. The list goes on, and on, and on, and on.
The combination of an extraordinarily intelligent model (Opus 4.5), a harness that lets it work autonomously, maintain a pseudo-memory, and execute commands (Claude Code), and complete browser control (Claude in Chrome) results in something truly powerful. In a very meaningful sense, Claude Code is a knowledge worker.
Yes, it’s closer to “someone smart on their first day of work” than “an expert who’s been doing this for decades.” But I am willing to bet that it can nonetheless do significant portions of your job.
This is a step change in AI. The much prophesied “agentic” era of AI is no longer theoretical. METR estimates that Opus 4.5 can sometimes do tasks that take humans almost five hours. I’ve noticed similar, if not more, impressive results in my own work.
Despite its capabilities, however, Claude Code and OpenAI’s Codex, its closest competitor, will likely not break into the mainstream in their current form. Unlike ChatGPT, there is a significant barrier to entry: you need to know how to use your terminal and be comfortable with using a handful of basic commands. This isn’t hard, but it’s intimidating enough for novices to be a problem. While it exists, this small barrier looks like the foundation of a two-tier economy. Those who learn Claude Code will see their productivity skyrocket. Everyone else will wonder why they’ve been left in the dust.
But you shouldn’t expect this gap to persist for long. Solving the accessibility challenge is a UX problem, and a fairly simple one at that. Anthropic and its competitors have already started building interfaces that make the underlying technology more accessible — Claude for Excel, for instance. And because Claude Code is writing much of its own code, its capabilities are likely to improve at a breathtaking pace.
That last point is, perhaps, the most crucial. Because while Claude Code’s generalist capabilities are impressive, the real gains will come from its coding skills — specifically, its ability to automate its own research and development.
This is beginning to happen. On December 27, Boris Cherny, the lead developer on Claude Code, said “in the last thirty days, 100% of my contributions to Claude Code were written by Claude Code” — 40,000 lines of new code. In the coming months, then, we should expect Claude Code’s already remarkable capabilities to improve extremely rapidly — with those improvements coming, in large part, from the system itself.
I’m sure many of you will have rolled your eyes reading this. The tech world is used to thinking about code expansively, as the ability to get a computer to do pretty much anything it is capable of. For everyone else, the true scope of what the ability to code enables can still feel remote. In the end, the only real way to “feel the AGI,” as they say, may be to try it for yourself.
If you do, I suspect you’ll come to a similar conclusion as me. Claude Code is not AGI. But for the first time, I can viscerally feel that truly transformative capabilities are within touching distance. The hypotheticals that underpin much of the AI policy debate — mass unemployment, catastrophic risks, the end of knowledge work as we know it — no longer feel hypothetical. The question is no longer whether AI will transform knowledge work. It’s whether anyone outside a small technical elite will notice when it does.
Wonderful to know that this is all possible, but as someone who's currently trying to transition into operations, I'm concerned that this would wipe out the bottom rung of that career ladder. I'm assuming that the solution is to learn Claude Code, but I don't know how to code. Does anyone have any resources they could suggest so that I can upskill and stay ahead of the curve?
Great article! The two-tier economy won't be about who learns the terminal. It'll be about who can specify intent clearly. The bottleneck is shifting from "can you execute" to "do you know what you actually want." That's a much stranger skill gap.