Two weeks. That’s how long I lasted.

I signed up for Claude Pro with the same excitement I get every time a new tool shows up promising to make my day easier. And for about 5 days, it delivered. Then the honeymoon ended on a Tuesday, mid-task.

First, the Love

Claude is genuinely good. Better writing than most of the alternatives, better reasoning on anything with nuance, and it doesn’t talk to me like a customer service script in a foreign country. For research, drafting, and thinking out loud on a business problem like analyzing WordPress export types or the different requirements for Lady Bird Deeds in each state, it’s one of the best AI chat interfaces I’ve used.

I wasn’t shopping for a replacement. I was happy. That’s an important detail, because this isn’t a “the product is bad” post. It’s a “the product doesn’t fit how I actually work” post. Those are different problems, and only one of them gets fixed with a feature update. (Attention: Anthropic please note if you are crawling this post.)

Then the 5-Hour Window Showed Up

Here’s what nobody tells you before you hand over your credit card: Claude Pro runs on rolling 5-hour usage windows. Use enough of it, and you get cut off until the window resets. Not slower, or lower priority. Cut off.

I found this out the way you find out about most limits: by hitting one. Mid-task, mid-thought, in a state of flow on really complex work. The kind of moment where you’re seven messages deep into finally getting the output right, and the tool just… stops helping you.

For a subscription I’m paying for specifically so I don’t have to think about limits, that’s a rough way to learn there’s a ceiling. It’s not that 5 hours is an unreasonable amount of usage. It’s that the window doesn’t care what you’re in the middle of. It just runs out.

My Beef with the Rolling 5-Hour Claude Usage Window

Some days I use Claude a little, some days I’m doing in-depth work that requires a bit more heavy usage like coding or analyzing data, and other days I don’t use it at all (because I like my weekends).

My work day is typically 6 to 8 hours. If I’m using Claude that day for in-depth, flow-state work, I want to be able to conclude that work and not be interrupted by artificial limits.

Claude purposely prevents you from doing that which completely interrupts your day and workflow.

I’m paying a monthly subscription and expect to be able to use what I’m paying for when I want to use it.

Instead they should give you max usage in a 7 day or 24 hour time-frame and let you decide when to actually use it.

The Workaround (Which Should Tell You Something)

Here’s the part where I turned into someone I don’t love being: a guy who wakes up early to trick software into giving him more software.

The move is simple. For a day where you are going “Claude heavy”, fire off something small and pointless first thing in the morning, like “what’s the weather today” at 5:59 am, and that starts your 5-hour clock as the 5 am window. This gives you from 5 am to 10 am (where you’ll use almost nothing until you start work at, say, 8 am).

Now you have 2 solid hours of heavy usage and a reset at 10 am for 2–3 more hours of heavy usage before lunch.

In turn, it resets again at 3 pm for any last minute Claude tasks between 3 and 5 pm.

This trick burns almost nothing, but it opens the window early. Do that before the workday starts, and you can squeeze out a third window before the day is done instead of getting stuck on window two right when you need window three.

It works. I used it. But let’s be honest about what “it works” means here. I was setting an alarm to send a throwaway message to an AI so that, hours later, I wouldn’t get locked out during actual work. That’s not a feature. That’s a workaround for a limit that shouldn’t exist in the first place, and once I noticed I was building a small morning ritual around it, I knew I wasn’t going to stick with this for long.

The tools I pay for are supposed to remove friction from my day, not add a 5am chore to prevent friction later.

If you are really fancy, you can automate this annoying 5 am task.

But I’m a man of principles and this one seems ridiculous.

What I’m Using Instead

I’m not saying I’m not going to use Claude but I am not going to renew my subscription when the time comes. ChatGPT and Codex have really improved with v5 models and have never given me the artificial 5 hour emergency brake with my OpenAI Plus plan.

But for Coding and heavy activity I’ve decided to use an agnostic AI tool called Cursor.

The difference isn’t that one model is smarter than the other in some abstract sense. It’s that Cursor lets me switch models on the fly.

Hit a wall with one model on a tricky problem? Swap to another and keep moving. Max out usage on one? Switch and keep moving.

There’s no morning ritual required to keep working through the day.

That flexibility is the whole thing. I don’t need one model to be perfect. I need to never be stuck waiting on one model to get in a state of flow again.

I don’t smoke or vape so I don’t need these “cigarette breaks” throughout the day, but Claude’s 5 hour window made me consider starting.

Cursor though, is fantastic for the ability to switch models if nothing else.

The Lesson Here

This isn’t really about Claude being bad, because it isn’t. It’s about a pricing and access to a model that assumes usage in bursts when a lot of us doing real work during business hours use these tools continuously.

A 5-hour window is a fine design if you use AI like a snack. It’s a bad design if you use it like a tool that’s supposed to be sitting open on your second monitor all day to help offload deep or analytical work.

The companies that win this next stretch of the AI tool wars won’t necessarily have the smartest model. They’ll have the access model that matches how people actually work.

Right now, that’s not what I’m getting from Claude Pro, and it’s exactly what I’m getting from switching between models in Cursor instead.

So I’m not renewing. Not out of anger. Out of the same logic I’d apply to any other subscription that makes me build a workaround around it instead of just working when I need it to.