Post #4164486

@huronbikes@cyberplace.social I'm not that far away from where @bagder@mastodon.social stands on this, in regards to security related findings. In the OpenVPN project we've received both slop and very accurate reports. Lately the reports has definitely had better and more plausible scenarios than in the earlier days. But it's also been a shift in who sends reports these days. Now it seems to be more frequently by users who actually knows how to write good prompts and does their due diligence in reviewing the reports before sending a report. Most of the noise we see now is definitely from people being too lazy to verify their reports. Or even test the reproducers their LLMs has created.