Despite all this, I have now noticed that any request that goes beyond a quick text answer requires solving a captcha [0]. Additionally, even after solving one and getting output (that has become less reliable in my personal test on top), I get another one before any additional request gets processed. Previously, OpenAI has enforced a Cloudflare captcha for accessing the page, either during high traffic times or when accessing the site via a public VPN, but never directly after a request.
I wanted to ask whether any of you have experienced something similar. For me, at least, captchas are a massive nuisance. They break the flow I like to use LLMs for (back and forth bouncing of ideas), are an accessibility nightmare, and are generally barely acceptable for free websites that need to limit fraudulent traffic. On a paid service with a reliable, long-paying customer, it is frankly insulting and makes switching fully to Anthropic more attractive than ever, even though gpt-4o still does certain tasks in my field more reliably with less prompting compared to Claude 3.5 Sonnet or 3 Opus.
Am I the only one affected, or are you encountering something similar lately?
[0] https://imgur.com/S6qmCYy