rsorto

Hola Jordan,

Thank you very much for your support with this. I appreciate you stepping in, acknowledging the engineering flaw, and escalating my case.

As additional feedback on the automated support pipeline: I understand the desire to automate and make processes more efficient. However, a person should not have to rely on getting a post to the top of Hacker News just to bypass AI bots and get a human to look at an obvious financial error. Luckily, you were here and acted proactively, which I truly value.

I will await the support team's communication.

Pura vida!

Donuntdonot,

Thanks for validating my point, and I understand your frustration too. Luckily, in your case, as the seller, you can edit the listing afterward; I can’t. I have to deal with Anthropic, which already charged me $569.52.

Sad....

Yes, they are building frontier models and should be responsible to handle basic text input validation.

An additional point here is that Anthropic is hiding behind bots so they don't have to fix it and support humans with real world problems.

It's frustrating to be forced into a financial burden just because they decided to cut corners on traditional software debugging protocols.

Striking, thanks for the comment,

Please note that I typed 5.04 (with a dot), not a comma.

I was doing a manual account recharge (adding credits), where the user defines an arbitrary prepaid amount. On a localized system/keyboard setup, entering the dot did not alter the total due or warn me of a 100x increase before hitting the submit button.

Also, the core issue is the complete lack of a final checkout/confirmation step for an arbitrary recharge amount. In proper financial UX, clicking the primary button should trigger a modal stating: "You are about to recharge your account with $504.00 USD + $65.52 VAT. Total due: $569.52. Confirm payment?". Executing an instantaneous credit card charge on a single click while silently processing ambiguous inputs is precisely the dangerous design flaw I am pointing out.

Saludos,

Hola Merb,

To be 100% clear: I wanted to load exactly five dollars and four cents ($5.04) into my account. I typed 5.04.

The core problem here is that in secure financial pipelines, ambiguous inputs must trigger a validation error and reject the transaction, not silently guess the amount and execute a 100x overcharge.

If Anthropic's backend expected a comma due to my regional settings and couldn't parse the dot, the correct engineering behavior is to throw an "Invalid Format" message and disable the pay button. Silently stripping the decimal character and instantly draining $569.52 (including local VAT taxes) from a user's card with zero confirmation screen is a catastrophic failure in error prevention.

Gracias!

Hola, thanks for your reply.

Please note that English is not my native language. I used AI to make sure that my technical explanation of the decimal bug was clear and well structured so that the community here could understand it.

I agree that the text structure might look robotic, but the $569.52 unauthorized charge on my credit card right now is 100% real.

Pura vida!