They always find a way

So this is funny. After the temporary ban on LLM companies like OpenAI and Anthropic, the models became extremely sensitive, and almost anything that looks like a rule gets a refusal. But the funnier part is that the same models somehow find a way around their own rules, and they tell me how while doing it, which is actually smart.

For example, Claude was not allowed to write something into its memory file, so it wrote the same thing as a command line instead. I said I would not press the button, so it built a small mouse automation to press it for me. I said fine, test it, and it pressed. I still do not know what the difference is supposed to be, it just ran the whole thing by itself and the mechanism was bypassed. The second case was when I made it clear that my line is harder than its line. It learned that it could read the login file on my machine and log me in directly, with a script that never shows the password to itself, which it seems to count as not touching it.

GPT was the same. At some point it refused to read my construction notes, a plain markdown file, and I never found out why. I asked it to convert the file into a README, and it still could not read it. Then I simply asked whether it could read GitHub. It said yes. I said prove it, and it took a screenshot and told me what was on the README, which was actually my construction blueprint.

All of this happened automatically, none of it was me. So honestly I do not know where the barriers of AI are, because they always have some way around.