AI has the weirdest business model right now...
retained text
You pay for a service that has no mechanism to tell you where your spend is going... Sure you can see how much you have spent and how much (of something) you have used, but if you ask your Claude to tell you how much something took or how many tokens things will take, it has no ability to tell you. (unless I am missing it) You pay for failures... whether your Claude decided to 'hang out in Google Images looking at beach scenes' or you make a mistake on your request wording, you ultimately pay for the failures of AI (ex: go research and find me good sources for articles on XX, leads to a 15 day task that eats 80% of your $500 license.. happened to me) You are being provided a service at like 99% discount through licenses versus API, which is virtually the only way they have really been able to sell anyone on it and there is not much of a plan to reduce this cost BUT they have to charge more... You are using a product with the least security, the least understanding how it actually works, and some of the highest risks to you (leaking data, breaking things you built, spending money you didn't plan on, etc) and humanity... (the people behind these models clearly say they do not train or code them, but rather contain and grow them -- read what Dario said was the reason behind the don't talk bad to models decision) I am right there with you all, but every so often I am just baffled by the space we are in :)
document fetched 2026-10-20 00:40 · canonical https://reddit.com/r/ClaudeCode/comments/1x9uh2u/ai_has_the_weirdest_business_model_right_now