• 1 Post
  • 100 Comments
Joined 3 years ago
cake
Cake day: June 17th, 2023

help-circle
  • I would say yes and no. I don’t think it’s entirely horrible as a programming assistant. I used it just a week ago to perform a series of UI upgrades. I probably wouldn’t have learned anything doing it myself, just mindlessly following the upgrade guide until it’s all done. It would’ve taken me a day at least, between meetings and other distractions, but an agent did it for me, unattended, in about 5 min. I still verified functionality afterwards and fixed a couple tests.

    In some respects it kind of feels like the argument I had with a coworker over the use of Lombok (a tool for injecting common but often tedious coding patterns in Java). I was on the side of not using it because some of the patterns are important to understand and not understanding the implementation can lead to misuse of them. Eventually I decided it was a “necessary evil” and that using stuff like that could free me up for tackling the stuff that is completely unique and wouldn’t be found in any library. The fun stuff.

    I still hate data centers and AI revenge porn and AI scams and people that replace perfectly functioning tools and experienced workers with AI bots that aren’t nearly as reliable. But I don’t think it’s an inherently bad technology.


  • Technically yes, it would make it harder, but a house with unpickable locks isn’t impervious to entry. In my personal experience memory exploits aren’t the primary methods hackers use to gain access or run custom code. I think layers of protection are more effective at stopping actual damage from being done. Run custom code, but you’re still an unprivileged user. Elevate your access but you’re still in a sandbox. Break out of the sandbox but you breach memory allocation and the environment is destroyed and rebuilt. And all the while you should be tripping alerts.