📖 The Full Story
The $749 Billing Dispute That Exposed AI's Broken Promises
By Tony Greenberg • February 2026
Video Evidence: Credit Request Denied
What you're seeing:
Tony requesting credits for wasted tokens due to agent errors. Response: "We don't offer credits." This is the moment that sparked the movement.
Artifact #2: Manus AI Recommends 60,000-70,000 Credits
What this proves:
Manus AI itself acknowledged the wasted tokens. After reviewing the work breakdown (52,000 tokens used), the AI recommended applying for 60,000-70,000 tokens of productive work credit — excluding overhead from context management and repeated iterations.
This wasn't a customer complaint. This was the AI platform's own assessment that the majority of tokens were wasted due to system errors and inefficiencies.
Yet when I requested those credits? "We don't offer credits."
How It Started
I've spent 25 years building fair billing systems for the IT industry. I created the first SLA for colocation in 1996. I co-architected the 95/5 billing model that became the standard for burst capacity. I've negotiated $24 billion in IT transactions.
So when I got charged $749 for a service downgrade on Manus AI, I knew something was fundamentally broken.
Here's what happened: I was using Manus AI to build a complex project. The agent made repeated mistakes, wasting thousands of tokens. When I requested credits for the wasted usage, I was told "we don't offer credits."
Then came the billing surprise: $749 charged to my account. When I disputed it, I got silence. 11+ days without a meaningful response.
Timeline of Events
Day 1
Charged $749 for service downgrade. Submitted dispute through official channels.
Day 3
Requested credits for wasted tokens due to agent errors. Denied: "We don't offer credits."
Day 5
Escalated to support. Received automated responses with no resolution timeline.
Day 11+
Still no meaningful response. Decided to build this movement instead.
Why Token Pricing Without Dollar Values Is Unethical
I've spent 25 years watching industries use opaque pricing models to extract maximum revenue while minimizing customer understanding.AI companies are repeating the exact same playbook.
When you charge customers in abstract units—"tokens," "credits," "data bits"—without clearly showing the dollar value, you're deliberately making it impossible for them to budget, compare, or understand what they're actually paying for.
This isn't new. I've seen this pattern destroy trust in multiple industries:
Telecom: "Data Bits Per Second"
In the early days of internet billing, telecoms charged for "data bits delivered" without any clear pricing model. You'd get a bill showing "2.4 million bits transferred" with no way to correlate that to dollars.How do you budget when you don't know what a bit costs?
The counting methods were black magic—they'd measure however they wanted, send the bill, and customers had no recourse. Sound familiar?
AWS: Deliberate Complexity
Amazon Web Services pioneered the art of pricing obfuscation. "Compute units," "I/O operations," "data transfer zones"— a deliberately complex model designed to make cost prediction impossible.
The result? Companies routinely get bills 300-500% higher than expected. AWS creates a quagmire where you can't understand your costs until after you've already been charged. That's not transparency—that's a trap.
Mobile Roaming: "Data Units"
Remember when mobile carriers charged for international roaming in mysterious "data units"? You'd return from a trip to a $2,000 bill for "847 data units" with zero explanation of what that meant in dollars per megabyte.
It took $2.1 billion in refunds and regulatory intervention to force carriers to show real pricing. Why? Because opacity enables overcharging.
AI companies are doing the exact same thing.
"You used 52,000 tokens" means nothing without dollar context. Is that $5 or $500? You can't tell. And that's the point—when customers can't predict costs, companies can charge whatever they want.
The solution is simple and proven: Show the dollar value alongside every token charge.
"This task used 5,000 tokens ($12.50)" gives customers the information they need to make informed decisions. Anything less is choosing opacity over transparency—and that's exactly what this movement exists to fix.
The Bigger Picture
This isn't just about $749. It's about a pattern I've seen destroy entire industries.
In 2002, WorldCom collapsed under $180 billion in billing fraud. In 2012, mobile carriers paid $2.1 billion in refunds for "bill shock."In 2020, cloud providers saw $12 billion in disputed charges during COVID.
AI is repeating the same mistakes.
Gartner reports that 62% of IT leaders can't predict their monthly AI costs. Usage-based pricing without transparency is creating a new generation of billing disasters.
But here's the thing: I still love these products. Claude is brilliant. Manus is powerful. OpenAI is revolutionary.
That's why I'm writing this. Not to tear down AI companies, but to help them build something that lasts. To show them how to create trust through transparent, fair billing practices.
This Can't Continue
Join the movement demanding transparent, fair AI billing practices. Sign the petition and help us reach 100,000 signatures.