I use Claude every day — coding, writing, planning, problem-solving. It's genuinely become part of how I work. And I quickly realized that my AI usage limits were becoming a barrier to getting real work done.
Treating Claude like a chatbot was clogging up the AI's ability to think, causing it to grind through tokens at such a quick pace that my usage limit would be hit and my momentum would be gone.
So I started digging. What I found was both simple and kind of mind-blowing: Claude re-reads the entire conversation from the beginning on every single message. Every reply. Every file you pasted. Every code snippet. All of it, every time.
That means message 1 costs, say, 1 unit. Message two costs two units (1 and 2). But the cost is triangular and as you get up to 20 messages it isn't 20 units...it's 1+2+3+...+20 = 210 units. And I thought, I need to tell someone about this.
I mentioned this to other Claude Pro users. Most of them had no idea. There was no dashboard showing it, no warning when a conversation was getting expensive, no explanation in the onboarding. You just suddenly ran out, confused about why.
So I built Tangerine. Not to sell anything. Not to pitch a premium tier. Just to make the invisible visible — a simple, private, browser-only tool that shows you exactly what your prompts cost, what your session overhead looks like, and how to write more efficiently so you actually get to finish what you started.