It’s easy to fall down the rabbit hole that is the hype surrounding Anthropic’s code agent Claude Code, a hype that really took off during the Christmas holidays and — at least in tech circles — is reminiscent of ChatGPT’s arrival three years ago. Claude Code, already being called both “the new ChatGPT” and “the end of SaaS,” and is considered by many to be the next big step in AI development and AI use.

If, like me, you spend a little too much time online, Claude Code has started to completely dominate your feeds lately. Of course, it “helps” that the algorithms give me “more of what I want,” but both LinkedIn and TikTok are overflowing with people talking about the excellence of Claude Code. AI influencers, ordinary developers, curious IT professionals, and, of course, a steady stream of obscure individuals who want to cash in on all the buzz.

People are vibe-coding everything from entire websites to small personal apps and finding new solutions for their workflows. Ethan Mollick, a professor at the Wharton School of the University of Pennsylvania, wrote about how he had Claude Code build an entire startup. In other circles, there’s talk of Gas Town, an orchestration solution that allows 20 to 30 agents to work in parallel. In San Francisco, it is said, people are letting “Claude swarms control their lives.”

In the last week, Clawdbot (now Moltbot) also jumped on the hype train; it’s a Claude Code-like AI assistant that you give full access to your digital life, if you dare.

It’s all fascinating and often incredibly impressive. Code is seen as the primary use case for generative AI (genAI), and when you see what Claude Code can do, it’s not hard to understand why. I haven’t tested Claude Code myself, even though more than one person has told me that I simply “must.” And I’m not particularly keen to do so (though I will probably get around to it later for purely educational purposes). Let me explain my thinking.

For starters, I’m not a developer and I can’t write code. Now, the point is that Claude Code is supposed to write the code for you, but it’s not quite that simple. The entry barrier is still too high for me, and the road to the really cool stuff is too long. I have a basic understanding of the processes and functions involved in software development, but my knowledge is far from enough.

Anthropic has begun to address this by releasing Claude Cowork, which is described as Claude Code for people who are not developers. That sounds interesting. But here’s the problem: what I’m being asked to do with it, how it’s being sold — it just sounds…boring. For some reason, the first example often highlighted for tools like this is that they can “sort your files on your computer.” That might be a good thing to do, once, if you have a lot of unsorted files on your computer. But it’s hardly a use case that’s worth buying and learning a new tool for.

If you look at what enthusiasts are doing with Claude Code, a lot of it focuses on personal productivity and “optimization,” both at work and in everyday life. That’s something I’m completely uninterested in. Sure, I also read “Getting Things Done” and “The 4-Hour Workweek” 20 years ago, but what I learned above all else was that the deeper you sink into that self-help swamp, the greater the risk you’ll end up spending all your time finding new ways to optimize things. (It’s also sad, but perhaps telling of modern working life, that people need to come up with “hacks” to be able to do their jobs.)

The main reason I’m meh about Claude Code is this: I lack the software perspective. This realization occurred to me recently when I read tech writer Jasmine Sun’s experiences with Claude Code. She points out how people who do parkour over time learn to see a city in ways that the rest of us don’t; they develop a “parkour perspective” where walls and stairs become something completely different from walls and stairs. In the same way, she suggests developers develop a kind of software perspective where all problems can be translated into, and solved by, software.

An ordinary user like me does not have software-related problems. I do not automatically see how something I do can be automated or optimized by a bot. So when I hear that Claude Code can solve my problems, I cannot think of a single one.

I think this perspective says a lot about the use of AI in general and isuseful to keep in mind. That’s true for both super users who hype the new solutions for people who “can’t keep up,” and for companies that are feverishly trying to get their staffers to embrace AI. I don’t think anyone is against a new tool that makes them more productive and facilitates their work, as long as they know what to use it for.

Perhaps companies’ AI training courses should focus less on learning specific tools and more on learning to identify the problems they’re supposed to solve, perhaps even developing a software perspective.

This column is taken from CS Veckobrev, a personal newsletter with reading tips, link tips, and analyses sent directly from editor-in-chief Marcus Jerräng’s desk, each Friday. Sign up for a free subscription here.