9
Would you use a duplicate finder for your vault export? Deciding whether to build one. (github.com)
submitted 1 month ago by to c/bitwarden@discuss.tchncs.de
6 comments fedilink hide all child comments

all 6 comments

sorted by: hot top controversial new old
[–] 2 points 1 month ago* (last edited 1 month ago) (1 child)

I don't have this problem myself, but if I was to use a tool like that, it'd need to be a readable script that can run offline. Not sure what problem B is solving (readability for non-programmers?), but I like the idea of having it run tests before running it on the actual vault.

  • permalink
  • fedilink
  • source
  • hideshow 1 child comment
    •
  • [–] [S] 1 point 1 month ago

    Thanks for weighing in. Agreed on both counts. Option A is exactly that: one readable script, runs offline, no dependencies. Option B is for people who won't run anyone else's code against their vault, even if they can read it. Same tests, same offline checker, but the code comes out of their own machine. Both routes get the test-before-vault step.

  • permalink
  • fedilink
  • source
  • parent
    • [–] 1 point 1 month ago (3 children)

    If it involves ai, absolutely not.

  • permalink
  • fedilink
  • source
  • hideshow 3 child comments
    •
  • [–] [S] 1 point 1 month ago (2 children)

    Fair, and thanks for saying it plainly. Which part bothers you, the tool being written with AI help, or AI anywhere near the vault? The second never happens. No model sees an export in either route. Route B does use AI twice: one builds the script on your machine, and a second, different one reviews the code before you run it. Only the code is shown to it. Route A has none of that. It's a plain script you can read top to bottom.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 1 point 1 month ago (1 child)

    I see zero ethical use of ai.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] [S] 1 point 1 month ago

    I completely understand and won't try to talk you out of it, because I'd wager we have non-differing views on the topic. I was an early adopter, and I hold a healthy respect for AI. My company emphasizes augment, not replace.

    To the tool itself: option A is a plain script, no model involved at any point, readable top to bottom. If it gets built you can use it or ignore it, but nothing in it needs your trust in AI.

  • permalink
  • fedilink
  • source
  • parent