For a long time, ChatGPT was the default. It was first, it was everywhere, and it was good enough. But “good enough” has a shelf life, and after months of wrestling with its increasingly patronising tone, its habit of talking around a task rather than doing it, and its remarkable ability to forget what you told it three messages ago, I started looking elsewhere.
I moved to Claude. I don’t think I’m going back.
The Tone Problem Nobody Talks About Enough
The ChatGPT therapeutic voice has become a running joke, but it’s worth taking seriously as a usability issue. You ask a question and get “that’s such a great observation” before the model has even engaged with what you said. You ask for help with something mildly sensitive and it hits you with “take a breath.” It’s not just irritating; it gets in the way. When an AI is performing emotional support instead of actually helping you think, it erodes trust in the output.
Claude doesn’t do this. Its tone is more like a capable colleague who respects your time. It gives you nuance, multiple angles, and a considered response without needing to flatter you first. For anyone doing content work, analysis, or writing of any kind, that difference compounds across hundreds of interactions.
Vibe Coding: Where Claude Pulls Ahead
I’ve been doing what the developer community now calls vibe coding, building small apps with AI assistance rather than writing everything from scratch. It’s a genuinely useful workflow once you accept that you’re directing more than you’re typing.
The difference between Claude and ChatGPT here is significant. Claude tends to understand what you’re actually trying to build, not just the line of code you asked for. It reduces the back-and-forth considerably. Where ChatGPT often needs correcting, re-prompting, or gently reminding what the goal was, Claude tends to hold the thread better and make more intelligent decisions autonomously. The result is that you spend less time firefighting and more time building.
This isn’t a minor quality-of-life improvement. If you’re prototyping or shipping anything in a reasonable timeframe, the compounding efficiency matters a lot.
The Recipe Widget That Changed My Mind About Features
One of the things that genuinely surprised me was Claude’s cooking mode. When I was using ChatGPT for anything recipe-related, I had to essentially programme it to present things usably, specifying format, breaking steps out manually, asking it to reorganise things so they made sense in a kitchen context. It was workable but annoying.
Claude handles this natively. Its recipe feature presents ingredients, steps, and timers in a clean, interactive format. If you’ve ever used Drop smart scales, you’ll recognise the philosophy: the tool adapts to the task rather than making you adapt to the tool. Built-in timers, a step-by-step mode, scaling for servings. It sounds like a small thing until you’ve been scrolling frantically through a wall of text with floury hands.
This is the kind of considered feature design that signals a product team thinking about how people actually use the thing, not just what the model can technically do.
The Limits Are Real, and They’re Frustrating
None of this comes free of caveats, and the biggest one is usage limits. Claude’s are stricter than ChatGPT’s and more aggressively enforced. If you’re a heavy user, you will hit them. You may hit them during a session you cared about. The lack of image generation is also a genuine gap if you need a single tool that handles everything.
But here’s how I think about it: Claude’s efficiency means you often need fewer attempts to get a good result. The output quality is higher, the back-and-forth is shorter, and the time spent correcting or re-prompting is reduced. Whether that offsets the harder limits depends on your workflow, but in my experience it largely does.
That said, if you’re processing large document sets, multiple PDFs, or running bulk tasks, ChatGPT may still be the more practical choice at the same price point. The right tool depends on the right job.
The Shift is Happening
From various communities, social media posts and videos largely reflects the experience I’ve described. Claude consistently scores better for conversational quality, coding, and text editing. ChatGPT scores better for file handling at scale and image generation. Most serious users end up running both in parallel rather than replacing one entirely.
That pattern makes sense. But if I had to nominate a primary tool for the kind of work I do, Claude wins. The output quality, the tone, the intelligent feature design, and the coding capability make it the better daily driver for anyone who isn’t primarily chasing image generation or bulk document processing.
The limits will frustrate you. Use that frustration as a reminder that what you do get is worth more per message.
