Do I have to be loyal to GPT?

​Read time: 5 mins

Hi Reader,

If you've noticed more people talking about Claude. It's not just hype.

This issue is about what's actually changed and whether switching tools is worth your time and energy right now.

Most people won't act on any of this, though. Default behaviour is powerful. ChatGPT was the first serious AI tool many of us adopted, and the first-mover advantage is real. It takes effort to re-evaluate something that's "good enough." The problem is that good enough, when everyone else is using it, is not a competitive edge.

From the experts

The biggest practical difference for me is that ChatGPT agrees with everything. You pitch an unfinished idea, and it calls you a visionary. Claude doesn't do that. It pushes back and asks clarifying questions in a user-friendly way. For anyone using AI as a thinking partner rather than a text generator, this matters a lot.

The outputs are also more usable in practice. I tested both on the same brief (setting up a forecasting pipeline Excel sheet), and the difference in what came back wasn't subtle. Claude's output was closer to something I could actually hand to my accountant.

Claude Cowork is a feature worth mentioning, though I'm still exploring it. What I've used so far (reorganising project files across a client's video production workflow) has saved real time. The OpenAI equivalent exists in Codex, but it's buried and badly explained. Packaging matters.

The Anthropic Academy is also well structured and easier to navigate than OpenAI's documentation, especially if you're still building your foundations.

Then there's the positioning side. Anthropic's messaging (human-centred, no ads in responses, explicit about what they won't do) is landing at exactly the right moment and adding salt to the wounds of its competitors.

There's something else worth saying here that I haven't seen many people acknowledge: tool choice is becoming a form of professional identity. In a meeting, saying "let me ask Claude" signals something different from what it signalled six months ago, and very different from asking GPT.

That said, Claude isn't perfect. Context management across long sessions can be frustrating. And the switching cost of rebuilding workflows and prompts is real, not something to take on casually. Though if you are considering the switch, try this prompt and guide.​

So, my position: I don't owe loyalty to any tool. ChatGPT served me well for over a year. Claude is serving me better right now. I'll follow the quality, not the brand.

What I'd suggest: if you haven't tried Claude yet, start with a task where you genuinely want pushback. A strategy brief, a positioning question, something where yes-and is dangerous. That's where you'll feel the difference fastest.

Would love to know your thoughts.

Escape the algorithm (Further reading and material)

NEWS: Meta and Google are found liable in landmark social media addiction trial. The jury took more than a week to make its decision. It's being seen as a major step in protecting children from online harms. The case is being compared to those against Big Tobacco last century, where companies were accused of hiding information about the harms of cigarettes. This topic warrants its own newsletter, but thankfully, it's all over the news.

What do you think of the result? Are the platforms really to blame — or the individuals?

DATA: Australia banned social media for kids under 16. Just 3 months in, here are the results.​

If you got this far,

you're part of 71% of subscribers, so thank you, Reader.

I've got some case studies up from recent work, if you want to take a look.​

If a section has particularly interested you, click the link of the corresponding section. It helps me prepare future newsletters.

If this was forwarded to you, you can subscribe here.

See you in two weeks.

​Peace,

Has

​​​