Login
You're viewing the federate.social public feed.
  • Aug 6, 2026, 3:46 PM

    Here's a tip (should be obvious, but apparently not) for companies who think they're making their documentation more "accessible" by feeding it all into an LLM and connecting a chatbot to the LLM to answer user questions…
    The accuracy of the chatbot's answers is entirely dependent on the comprehensiveness and accuracy of the documentation you train it with.
    If your docs are incomplete or have a ton of old, inaccurate articles, your chatbot will be shit.
    Looking at you, .

    💬 2🔄 4⭐ 3

Replies

  • Aug 6, 2026, 3:50 PM

    @jik

    "Is your AI _teammate_ not being as helpful as you'd like? Why not ask it the same mf'ing question fifteen times to see if we've cleverly hidden the information you require behind some weird gate that only occasionally bubbles up to the top-layer of the AI slop-stack we've created?"

    👀

    ... I hate it.

    💬 0🔄 0⭐ 1
  • Aug 6, 2026, 5:22 PM

    @jik Sadly we're also saddled with Atlassian at $dayjob but I avoid Rovo like the damned plague.

    💬 1🔄 0⭐ 0
  • Aug 6, 2026, 5:49 PM

    @pertho You may want to mention this to your security team in case they're not already aware of it:
    federate.social/@jik/117049302
    At my company after finding out about this yesterday we made the decision to turn off Rovo until Atlassian addresses this issue, and even after the issue is addressed we aren't going to turn it back on unless we hear from people that having it turned off significantly impacted their productivity.

    💬 2🔄 0⭐ 1
  • Aug 6, 2026, 10:16 PM

    @jik I'm going to tell them first thing tomorrow morning. 👍 Many thanks for the heads up!

    💬 0🔄 0⭐ 0
  • Aug 8, 2026, 5:42 AM

    @jik Great news! They never turned it on. 👍

    💬 0🔄 0⭐ 1