claude-code - 💡(How to fix) Fix Your pricing model is predatory and your product does not deliver on its promises

Preflight Checklist

• I have searched existing issues for similar behavior reports
• This report does NOT contain sensitive information (API keys, passwords, etc.)

Type of Behavior Issue

Claude modified files I didn't ask it to modify

What You Asked Claude to Do

I paid for a full year subscription to Claude Code. I also pay for GitHub Copilot Pro+ and OpenAI Codex Plus. I run these tools side by side. I know what I'm talking about.

Claude Code is a scam — not because it doesn't work at all, but because of how it's designed to consume tokens. Small, routine tasks burn through usage at a rate that is clearly engineered to drain subscriptions, not to serve developers. This isn't incompetence. It's architecture. And that makes it deliberate.

You market this as a professional developer tool. Professional developers notice when a tool is optimized against their interests. We notice when usage metrics don't match output value. We talk to each other.

I don't think the engineers building this are proud of it. I think decisions made above them turned what could have been a genuinely useful tool into a billing machine dressed up as an AI assistant.

Here's what I want you to know: I'm not just angry about my money. I'm angry because you are poisoning trust in AI tooling for developers who are genuinely trying to build things. You had a chance to win long-term loyalty. You chose short-term extraction instead.

Fix the token consumption model. Be transparent about what tasks cost before they run. Or keep losing customers who will tell everyone exactly why they left.

— A developer who won't be renewing

What Claude Actually Did

What I observed:

I used Claude Code for a few basic feature implementations — the kind of routine tasks I regularly perform with GitHub Copilot Pro+ and OpenAI Codex Plus.

The difference in resource consumption was immediately noticeable. With other tools, I can work through multiple tasks without hitting any usage ceiling. With Claude Code, basic feature work visibly and rapidly drained my usage quota.

I am not a light user. I run these tools side by side and I know what normal consumption looks like for this type of work. What I observed with Claude Code does not match the output value delivered. The tool re-reads files repeatedly, spawns unnecessary tool calls, and generates verbose intermediate steps that appear to serve token consumption more than they serve the task.

The core problem:

This is not a usage pattern issue on my end. Other AI coding tools handle the same workload at a fraction of the cost. If Claude Code burns significantly more usage on equivalent tasks, either the architecture is inefficient or it is designed this way. Neither is acceptable for a paid annual subscription marketed to professional developers.

Expected Behavior

What I want:

Transparency on what a task will cost before it runs A meaningful audit of whether tool call frequency is optimized for the user or for billing A straight answer on why equivalent tasks cost dramatically more here than on competing platforms

Files Affected

Permission Mode

Accept Edits was ON (auto-accepting changes)

Can You Reproduce This?

Yes, every time with the same prompt

Steps to Reproduce

No response

Claude Model

Sonnet

Relevant Conversation

Impact

Critical - Data loss or corrupted project

Claude Code Version

2.1.133 (Claude Code)

Platform

Other

Additional Context

No response