@theo I genuely don't know how people say Codex is better than Claude Code. I've had both ($200) for a couple of months now, and even a reset every 2 days in Codex wouldn't be enough to convince me to stay with Codex anymore. Talk about why people are wasting their money.
@bcherny@tt_tisht@ipv4fan It would be great to have better control over auto mode. When I haven't released a project yet, I want it to have access to the whole environment, no questions asked.
@bcherny@tt_tisht@ipv4fan I love auto mode, but it's sometimes annoying. Even if I specify in the conversation that I allow it to merge or connect to the database, it'll reject that anyways. I have to turn on manual mode for the model to do a certain thing.
@OfficialLoganK Is there a way to report false positives or get support for this? The filter is too aggressive for any app that deals with family-friendly content.
Similar reports: https://t.co/bsnPdyLrlW
Would appreciate any guidance on how to escalate this to the safety team.
@OfficialLoganK These are normal, everyday scenarios. The CSAM filter has so many false positives it's blocking legitimate use cases.
We've tried every prompt engineering trick - removing ages, using "young person", rewriting everything. Nothing works reliably.
@OfficialLoganK Be better at making the output just as long as asked. If you ask for a story, even if you tell it to make a long story, the length is a bit random.