Over the last 8 years, I have been a software developer. In this line of work, when you are stuck, a bug you can't fix, an architecture decision that goes round in circles, you call a colleague. You ask in tech community forums. You scour Stack Overflow at midnight. It gets the job done, eventually, but it is a great time consuming. Always time.
When ChatGPT appeared, it made me feel a thing I had never expected: relief. There was something that could come to me halfway, instantaneously, without a single word of back-and-forth replies. Yet, the relief only lasted for a short while. ChatGPT looked good at the first glance, but not reliable in the actual work, especially when it is the kind of detailed, technical work that I require. So, I did what engineers do: I kept on experimenting. DeepSeek. Perplexity. I have tried a few others. Each one brought something, but none of them were an exact fit.
Next I gave Claude a shot. The first time it really seemed like the tool was able to keep up with the depth of my inputs. So I kept pushing, running my hand over the wall suddenly came up. Usage limit reached. Please try again later.
Maybe I thought: okay. That's what free plan means, having limitations after free usage. That's the way it is. So I upgraded to Pro.
The Pro Plan Promise and What Actually Happened
Moving up felt like a big step out of the pressure cooker. More time, more doors, less getting in each other's way. That's what the upgrade was all about, right? The shiny upgrade kept that promise for the first few days, but then I stumbled upon a surprise.
Even with the Pro version, the restrictions were there but cleverly disguised. A daily limit here, a weekly maximum there. After relying on Claude for a couple of really intensive days, I would have to wait until the following week to be able to use it again.
Paying and still being cut off, that feeling, is a very specific kind of aggravation. It's not the same as when you run out of phone data and that feels like a fair trade-off. This was a lot more intimate. The tool seemed to have made a judgment on me that I was using it too much.
I was paying for access but apparently, there was a limit to how much access I was allowed to access
So I decided to do what any other sane human being would do: I checked the higher tiers. And there it was, the scheme was laid out in front of me. The more you fork out, the more you get to play. Not a standard product at a standard price but a making-up-the-price-your-budget level.
That was the time I thought of it and I just couldn't get rid of the thought: Could it be that they are exploiting us?
Let's Talk About What's Actually Happening Technically
In order to be fair both to Anthropic and to myself (because I wanted to be clear about what I was agreeing to), I went and found out why those limits were put in place originally. And there is an authentic explanation but it's only part of the story.
The Technical Reality
When you send a message to Claude, it gets divided into small parts, called tokens roughly 3 to 4 characters each. Every word you write, and every word Claude replies, uses tokens.
Complex queries lengthy documents, in-depth analyses, multi-step reasoning take a lot more tokens compared to simple questions. These operations require high-end GPU hardware, considerable amount of electricity, and large-scale server infrastructure. It really costs money, on a large scale, every single time.
Usage limits, theoretically, exist to distribute that cost among millions of users. They are not arbitrary they reflect real resource constraints. However, the way those limits are set is a business decision.
So, yes the limits have a technical basis. But the problem with that explanation is that, it does not really explain: why are heavy users, i.e., the ones who get the most value from the tool, always the ones who hit the limit? And why does the limit conveniently go up each time you pay more?
The Uncomfortable Pattern
AI isn't just the novelty we marvel at when we think about it anymore, it is now the very infrastructure of the world at large. People employ such systems in their work, business, and lives. Dependency on AI is not some fancy talk, it is undeniable and soaring exponentially. Just imagine you are doing a project and suddenly the tool stops you, what comes to your mind, rationally that it is server cost? Well, maybe not. Most probably, you just think: I want to get a better plan.
That inherent tension, that 'I need to get the tool instantly' kind of urgency, is actually the very kind of friction which turns free users into paying ones and payers into higher-tier users. Whether that is by intentional design or only a by-product of tiered pricing, the result is one and the same.
I'm not implying that Anthropic with their intentions is kind of skeptical about that. Operating highly advanced AI models is not only expensive but far more requiring an overhead of funds to keep it running. But there is also a second side to this story: the difference lies in whether the price is set just to cover the costs or if it is geared to take advantage of the urgency of the situation. Unfortunately, at present, I am rather uncertain regarding which one is the current pricing model.
So, Are They Using Us?
Mostly, yes. And in fact, we are also using them. That's the truthful response.
The AI field has rightfully recognized that folks are willing to pay for tools that really help them save time, lessen their efforts, and enhance production. No, this is not exploitation after all, this is a value exchange. However, if the user experience is designed in such a way that you find yourself stuck exactly at the phase when you are most dependent on the tool, then it stops being a service and it becomes a subscription trap.
My suggestion? Don't upgrade just out of frustration. Instead, consider whether you are upgrading because the higher level really fits your requirements or because the limit was reached at the time when you were most desperate. These are completely different reasons, and only one of them is an appropriate ground for a money decision.
And regarding Claude I am still working on it. It's still very much okay. However, I have now changed my attitude to a more wary one, having realized that in the AI era, even having the access has become a product. And like all products, its price is determined by what the market can withstand.
Comments