pull down to refresh

ChatGPT Agent casually clicks through “I am not a robot” verification testarstechnica.com/information-technology/2025/07/openais-chatgpt-agent-casually-clicks-through-i-am-not-a-robot-verification-test/
251 sats \ 2 comments \ @south_korea_ln 31 Jul AI
I've noticed those captchas and other safeguards becoming increasingly complicated and convoluted this year. Related?
As a side note, I haven't tried the agent option yet, but openAI is doing a lot more "thinking" since last week, even in the default o4 mode. The python and shell scripts it writes are usually overkill for the small tasks I ask it to do.
write
preview
related posts
0 sats \ 0 replies \ @optimism 15h
I've noticed those captchas and other safeguards becoming increasingly complicated and convoluted this year. Related?
I think so. I'm now often spending up to 2 minutes to access a cloudflare gated article or a youtube video due to the endless bus and bike identification. Since I don't hear too many people complaining about this thus far, it must be the VPN + tor and the fact I have all browser security settings up (cookie expire per tab, ad and 3rd party script blocking) and never log in anywhere. It is now faster and there is less friction to use less encouraged methods to for example watch video.
They're losing this battle but they're doing max damage to my most precious asset - time - in the meantime.
o4 [..] The python and shell scripts it writes are usually overkill for the small tasks I ask it to do.
I've been using mistral-small for this (it's non-reasoning) and it works fine. Qwen3 (original release from 2 months ago) also works well if you /nothink it.
reply
0 sats \ 0 replies \ @SimpleStacker 31 Jul
Haha this is the most ridiculous one i've encountered, and it was on Github of all places
#1007750
reply