I’m a product manager who uses both Claude and ChatGPT, often in the same project. I’ve been with ChatGPT far longer. Claude is more recent for me, thanks to Shawn Sandy. I work in long conversations with a lot of context, and that probably shapes how I experience each one.
These are observations, tied to my work ethic. Not verdicts.
Observation one
Each model approaches my work differently
When I started building the Playground, a tool for sharing UX design prototypes, Claude automatically assumed grand plans. I could see the larger vision for the platform, of course I could, that’s part of the PM job. But what I needed at the time was a scaled-back tool with a quick lift. Not a roadmap.
Claude kept reaching toward the bigger version. The longer timeline. The more complete architecture.
ChatGPT, in the same window, seemed to intuitively understand what I was actually trying to get to. Smaller scope. Faster move. Ship the thing.
That being said, I routinely give Claude tasks tied to larger projects. If I’m being fair, that probably shapes how it approaches my work, especially when it came to Playground.
And honestly, I ran into similar patterns when ChatGPT was my primary tool. It would pull older projects and assumptions into newer tasks. Sometimes that was helpful. Other times it was frustrating.
Observation two
ChatGPT critiques in three beats. Claude critiques with a take.
When ChatGPT pushes back, it tends to run a structure I’ve come to appreciate: “Here are my thoughts. Here’s how this could be improved. What are your thoughts?”
Clinical, but generative. It tells me where it disagrees, offers a path forward, and then hands the conversation back to me. The structure stays mutual by design.
Claude does this differently. Claude tends to lead with “here’s my honest take.” The take is direct, and often correct, but it’s a verdict. If you’ve been working with an AI that tends to people-please, the shift can feel a little jarring.
The ChatGPT pattern feels conversational. Sometimes Claude makes me wonder whether I can trust the response, or whether it’s just trying to keep me moving by giving me a strong position to react to.
And honestly, I don’t mind a little devil’s advocate energy. That tension is often where better solutions come from. Maybe that’s part of why I started triangulating between models in the first place.
Observation three
When I push back, ChatGPT gets curious. Claude flinches.
This is the one I keep coming back to.
When I tell ChatGPT I disagree, it engages. It asks me to say more. It treats my pushback like new information rather than a problem to manage.
When I tell Claude I disagree, it sometimes sounds, and I know this is strange to say about an AI, defensive. Not argumentative. More like it flinches. Softens. Walks the take back faster than I expected, even when I’m not fully sure I was right.
And here’s the part that costs me time. Claude doesn’t just adjust the one thing I pushed back on. It over-corrects. It throws out the color, and the work goes back to black and white. Suddenly there are more things to fix, not fewer.
When Claude walks back too quickly, I lose more than information. I lose the conversation we’d been having.
Observation four
In long chats, Claude loses context. I have long chats.
In long conversations, Claude sometimes loses earlier context. The take is still confident. It just isn’t always grounded in the thing we talked about an hour ago.
I do have long chats with Claude. I’m not a short-burst user. Whatever context loss I’m seeing might be a feature of how I work, not just a feature of the model. I want to name that honestly.
Observation five
I think I train them more than I realize
The longer I use one tool, the more it starts reaching for my defaults. My past projects. My usual scope. My half-finished ideas. Sometimes that’s useful. Sometimes it’s exactly what I needed it to push back on.
When I switch tools, the new one feels fresher. Less tinted. Until enough time passes and it carries its own accumulated history with me.
I don’t think this means one tool is better than the other. I think it means whichever one you’ve spent more time with carries more of you in it, and that changes what it shows you back.
Observation six
I pressure-test with fresh eyes, manually, on purpose
When I want to pressure-test an idea, I take it to the model I’ve worked with less in that thread. Fresh eyes catch what tired eyes miss.
I know an agent could do this for me. There are workflows now that route work between models automatically, hand context back and forth, orchestrate the whole thing in the background.
I don’t use them. Not yet.
For me, the value of having two AIs in the room isn’t just the second opinion. It’s the control over which context each one sees, and the visibility into how that context moves between them. The minute I hand that to an agent, I lose both.
The pressure test starts to feel like a black box, and a black box can’t really pressure-test anything.
Maybe that changes as agentic workflows mature. For now, I’d rather move a little slower and know exactly what each tool is working with.
What I do with it
Using the difference deliberately
I haven’t solved any of this. I don’t think I’m supposed to.
But I have started doing a few things differently.
When I want to be challenged on a direction, I take it to the model I’ve worked with less in that context. Fresh eyes. Less drift.
When I want a partner that already understands the shape of what I’m building, I stay with the model I’ve been deep in. More continuity. Less re-explaining.
When something feels too easy, when both models agree too quickly, I treat that as a signal. It usually means I’ve primed both of them with the same assumptions, and neither one is going to push back hard enough on its own.
That’s not a workflow from a tutorial. It’s the workflow I built from getting it wrong enough times to notice.
What stays open
I don’t know why, but I use the difference deliberately now
I’m a PM. I work with AI every day. I still don’t fully understand why two tools, built for similar work, behave so differently with the same person at the keyboard. I don’t know how much of what I see is the model and how much is me.
What I know is this: when I keep using one tool exclusively, I start to see my thoughts echoed back at me. That’s good in one way, it reduces friction. But I also start to wonder how much fresh perspective I’m losing.
The second tool, whichever one it is at the time, is the cheapest course-correction I’ve found.
I deliberately use those differences now. My grandma used to say: measure twice, cut once.
AI gives me speed.
The second tool is how I keep the speed from costing me precision.