I posted yesterday: "I love that people have to explain what 'unlimited' means. Maybe we shouldn't use that word."
The context was Claude Code pricing and what "unlimited" actually covers.
When Anthropic advertises "unlimited" Claude Code usage at $100/month, there's enthusiasm in the product marketing sense — yes, it's a good plan. But there's also immediate confusion in the customer sense — unlimited what? API calls? Runtime? Number of projects? Concurrent agents?
Because "unlimited" in software never actually means unlimited. It means "enough that we're comfortable saying unlimited for marketing purposes" or "until you hit a hard constraint we haven't publicized" or "we've priced it assuming X usage pattern and if you exceed it we'll notice."
I've been thinking about the pricing since I upgraded. $100/month is real money. Not venture-backed startup money. This is my actual side-project budget money. So I want to understand what I'm actually paying for.
From practical experimentation over the past two weeks:
I can run Claude Code continuously if I want. The agent doesn't seem to have rate limits in the way GitHub Copilot does (where I'd hit a wall after a certain number of requests per hour). But that might just be because I haven't tested hard enough.
The real constraint seems to be the token limit on individual prompts and the context window. I can't give Claude Code a 200MB codebase and ask it to rewrite everything. There are practical limits.
The time constraint is real too. A task that runs for 8 hours probably hits some kind of timeout, though I haven't tested that either.
But the word "unlimited" obscures all of that. It sounds like: "Use it however you want, as much as you want, forever." That's not true. There are limits. They're just not advertised precisely.
I'm not being cynical. I'm being practical. When you're evaluating whether to invest in a tool, "unlimited" tells you nothing useful. You need actual numbers. You need edge cases. You need to know: "At what point does this stop working?"
Having used Claude Code now for several weeks, I think the $100/month price is fair if you're using it the way I am: multiple projects, daily usage, letting it run long-running tasks. The limits I've found are practical rather than artificial. The tool doesn't seem to be metering me or restricting me.
But that's not what "unlimited" communicates. What "unlimited" communicates is "you never have to think about cost" and that's not quite accurate.
I'd rather Anthropic say something like: "Claude Code: $100/month for continuous use on multiple projects, typical context limits apply, no per-request fees" or even just "pay once, use as much as you need."
"Unlimited" is a marketing word that requires explanation. And once you need an explanation, it's failed.
From a customer perspective, I'm happy with the tool and the price. But from a communication perspective, the messaging could be clearer. And in a market where several powerful tools are competing, clarity wins.