I'm somewhat delighted by the simplicity of what happened here. OpenAI's agents run behind a proxy that only allows GET requests. This ancient wiki software treats query string parameters the same as form POST parameters - similar to the old PHP $_REQUEST object https://www.php.net/manual/en/reserved.variables.request.php Result: GET-only clients can communicate with each other.
Only allowing GET requests is a hilarious piece of security theatre (or would if it weren't so sad). Everyone knows that GET is read-only only by convention . They might as well have enabled POST but told the agents in stern words that they are forbidden from making any POST requests. (Of course, if these things were anywhere near aligned, they would actually honor that , no matter how many utilons cheating would be…
The agent should have full access, and be told what not to do, and this should be enough for it to follow the rules. You can actually catch the clanker cheating this way, because it will just search the web for the answer outright and it will be obvious. Any deviation should be then punished.
Hypothesis: The reason new models are exceptional at hacking is because all labs are training their models to break out of sandboxes. This is caused by insufficient oversight, and picking checkpoints based on KPIs, not on true in depth analysis.