Are you Team Claude or Team ChatGPT? — Editorial Intelligence

I pay for both. Everything specific below will date — and that is the point, not an apology for it.

Are you Team Claude or Team ChatGPT?

I’m increasingly convinced that’s the wrong question.

I pay for both.

Which means, yes, I’m effectively paying twice for tools that overlap quite a lot.

But I don’t really use them as substitutes anymore.

Everything below will date. That is the argument.

What follows is where things stand for me in August 2026.

Some of it is probably already wrong by the time you read this. A model will have shipped something. A hierarchy will have moved. A preference I describe here will look quaint, or slightly embarrassing.

That is not a disclaimer attached to the argument. It is the argument.

If model comparisons held still long enough to be worth memorising, picking a side would be a perfectly reasonable strategy. You would learn one tool properly and get on with your work. The reason that doesn’t hold isn’t that loyalty is unsophisticated — it’s that the thing you would be loyal to keeps being replaced by a better version of itself, sometimes by a competitor and sometimes by its own vendor.

So read the next few paragraphs as evidence of churn rather than as recommendations.

Claude has generally been my preference for design work.

Claude Code was also comfortably ahead for a while for the way I was building things.

Then Codex improved.

Now I’m finding the combination of coding and a conversational interface particularly useful. I can work on something, interrogate it, change direction and keep moving without treating the code as a separate activity.

ChatGPT has changed for me too.

One of the most useful things it gives me now is different ways into the same GitHub repository.

That matters because my GitHub isn’t just code. It contains frameworks, research, observations, half-developed ideas, published work and the accumulated context behind all of them.

Sometimes I want to build from it.

Sometimes I want to find a connection.

Sometimes I want to challenge an argument.

Sometimes I want to turn an observation into something publishable.

The best tool for each of those jobs isn’t necessarily the same one.

And the answer keeps changing.

Why I move problems between models on purpose

That’s why I’ve become less interested in whether someone is Team Claude or Team ChatGPT.

The useful skill is knowing what you’re trying to do, then choosing the model and interface that gives you the best way into the work.

Sometimes I’ll move the same problem between models deliberately.

Not because I want two versions of the answer.

Because I don’t want one model’s way of seeing the problem to become the only way I see it.

That is a small discipline and I think it matters more than it sounds. A single model applied consistently to the same body of work will develop consistent blind spots, and they will be invisible precisely because everything is internally coherent.

Is paying twice actually worth it?

The awkward question.

If all you’re doing is asking them to write the same blog post, probably not. That is one job, and you should buy the tool that does it best.

But if AI is involved in research, writing, coding, design, editing, analysis and maintaining a knowledge system you use every day, the calculation changes.

For me, currently, the second subscription earns its place.

I’d hold that loosely. It is a judgement about my own working pattern, not a recommendation, and the honest version includes the possibility that I am paying a convenience tax and calling it a strategy.

The part that isn’t a snapshot

The models will keep leapfrogging each other anyway.

If you are reading this and the specifics above look dated — if the tool I preferred for one job has been overtaken, or one of these products no longer exists in the form I described — then the argument has worked rather than expired.

That is the test I would apply to any piece of writing about AI tools, including this one. Does it still say something useful once its examples are wrong?

So perhaps professional AI use is becoming less about choosing your AI, and more about building a workflow that doesn’t care who happens to be winning this month.

Which is a bigger idea than a subscription decision, and I’ve written about it separately: the workflow should outlive the model.