You know that moment when a big launch drops, half your feed switches overnight, and your finger starts hovering over the cancel button on a tool you actually like? I was right there this week.
Then I found a post from a LinkedIn creator who did the unglamorous thing first: actually testing the new ChatGPT Work before telling anyone to jump. What the author found stopped me cold, and I think it saves you a real headache.
The feature list isn’t the problem
Let’s get the honest part out of the way. ChatGPT Work does what Claude Cowork does. Agents and all. The original poster confirms it directly, no hedging. So if you were waiting for someone to say “the features aren’t there,” that’s not the story.
The story is what happened next.
The wall nobody’s talking about
The author hit “You’re out of usage” in the middle of the first task. On a paid plan. Not an edge case, not some monster workflow. Just normal work, cut off halfway.
And it gets rougher. If you want more ChatGPT usage, you can’t buy it. There’s no $100 top-up tier like Claude has. Your only path up is Enterprise, and Enterprise starts at 150 seats minimum.
Read that again if you’re a solo operator or a small team. There is no middle step. You’re either capped, or you’re buying 150 seats.
Side by side
ChatGPT Work
- Agents and the full feature set: present and working
- Image generation: still the best available, better than Gemini by a clear margin
- Usage on paid plans: runs out mid-task
- Buying more usage: not possible, no top-up tier
- Next tier up: Enterprise, 150 seats minimum
Claude
- Cowork covers the same agent territory
- Image generation: not the strong suit
- Usage: a $100 tier exists when you need more headroom
- Scaling up: you pay for what you need, no seat minimum blocking you
Why most people are blaming the wrong thing
This is the part I keep thinking about. The creator points out that plenty of people saw the announcement and got ready to jump. Very few actually tested it. And when the work stalls, most of them assume their prompts are bad.
The prompts are fine. The plan stops you mid-task. Those are completely different problems, and only one of them is your fault.
That’s a fantastic catch, and it’s the kind of thing you only find by using a tool on real work instead of reading the launch post.
The one reason to keep ChatGPT
The author is clear about this: image generation. Still the best, by far. Better than Gemini. If images are a real part of your workflow, that alone justifies keeping the subscription, no matter what you decide about agents.
Worth noting: this industry pro also flagged plainly that nobody’s paying for the Claude enthusiasm. It’s just where the work is moving faster right now. I appreciate that kind of transparency.
The recommendation
The post’s author lands on advice that’s almost too simple, which is exactly why it works:
- Ignore the launch noise and the switching frenzy in your feed
- Pick one real task, something from your actual week, not a demo prompt
- Run it end to end on the new tool
- See if you can finish it before the plan stops you
- Decide based on that, and nothing else
Cancelling a subscription because of a feature announcement is a bet on marketing. Cancelling because of one finished task is a decision based on evidence. Big difference.
I’d add one thing to the author’s list: check whether the tool sells you more capacity when you grow. That’s an easy question to skip when you’re excited, and an expensive one to answer later.
The full post has more detail on the testing and where each tool actually breaks. Go read it on LinkedIn, it’s worth the two minutes.