Can confirm
Despite using Brave AI which revolves around looking the web and summarizing the info from it (It’s a good thing at first glance until you realize that even search engines can suck which can make it worse, especially when it gets cached data that doesn’t even exist if you click it). And I have trusted it with most things and called it out on it’s Contradiction and more.
On the one hand you had Brave AI correctly open source intelligence Bus Contractors to help with someone find missing bags since at the time we didn’t know who it was until someone confirmed it to us.
On the other hand you had Brave AI that lied (but not fully) that Malmal app, a collaborative drawing web app is open source, called it out on that and just said “[Well f you] my point stands, its no longer open source”. Alright brave WHY DID YOU IMPLY BEFORE THEN IT IS, I ASKED TO PROVIDE OPEN SOURCE WHICH YOU MOSTLY CORRECTLY DID AFTER FACT CHECKING (I should have known the inability to find the malmal’s github should have been the red flag, at least the good thing with being a web app is that its isolated in your browser and you can give minimal information)
Sorry I calmed down
What I later realized is that, Brave wasn’t technically wrong, Malmal USED To be open source in an old version before the new one where they closed and obfuscated the code so its not readable in web tools but it could also be wrong in that regard. What brave had a stroke on is the fact that its LLM was seeing a cached version of the Github page thinking its alive just archived, but when you click the link, it gives you a 404 error on the non cached version.
So yeah its not perfect and this is why im disgusted when people cant at least learn the skill that’s needed so that if LLM does make the mistake it always does and wraps it in being confidently correct you can correct it and if it insists, check yourself the sources first it gives then see if its right or wrong, if its still wrong point the factual part at it.