alt Hacker NewsI was going to cancel my gemini membership today ..... still going ahead. In my experience, gemini 3.1 pro, 3.5, 3.6 flash constantly lie too much about completing their tasks whereas sol (even though equally dumb) never claims something has been done when it hasn't been.
Replies
seunosewa • yesterday at 7:23 PM
Two thoughts:
1) It makes sense to try 3.7 flash before cancelling.
2) Prompting models to be honest is surprisingly effective in my recent experience. But only if they listen to instructions.
That's a good point and something I encountered yesterday. On a multi-agent task, Gemini was the only model that got near the end, ran tests, saw it had issues, took a screenshot, saw the issues, and then said, "I'll mark it complete" and delivered.
I was paying for Ultra, then downgraded to Pro and at this point I'm near abandoning it. Agy as a harness also has a tendency to constantly request significant elevations to perform routine operations.