Comments
2 comments
-
I appreciate the detailed feedback from a fellow developer. While Prompt AI's deep integration is our strength, we hear you on model choice and are constantly working to improve the core AI.
-
Thanks for the reply. Copilot is now in SSMS (honestly, I didn't think they were going to do this)…let's me pick my model…runs as a standard chat window like everything else…has my thread history during the session. It needs some work but it's already better than what I had. Again, my point of view and I realize that.
I hope you guys figure it out but right now…real danger of being eclipsed unless you can bring something majorly better to the table.
At this point, why would I use prompt AI?
Please understand - I love my Red-Gate tools and I was totally excited to see you guys were doing. I'm just being honest.
Add comment
Please sign in to leave a comment.
I really thing you guys are headed in a bad direction. After all this time, Prompt AI still doesn't return answers as good as Sonnet (or Grok or ChatGPT). I am not certain who's idea it was to attempt to compete with them.
It would have been soooo much more useful to let us pick our model.
This is, of course, my experience and I am not suggesting this proves anything - except to myself. I should mention I am full stack so I spend way more time in IDE's than SSMS.
I have taken the time to send exact same requests to Prompt AI and Sonnet. Sonnet is only looking at schema dumps. Prompt AI will either fail to give me a useful query or give me an inferior one, pretty much every time. I say inferior in that the Sonnet response seems to be more complete - for example, I asked both to examine a trigger for conditions that were closing a ticket and to provide a query that would detect where the conditions were true but the tickets were unclosed.
AI Prompt gave me a query that detected 93 records. Minimal info.
Sonnet gave me a query that found 145 (it was right) and WHY it was wrong (referencing the various conditions from the trigger). Again, same prompt used.
I basically have stopped bothering to use it - It also has some very aggravating habits - dropping code in wherever my cursor happens to be. Clearing if I open/close it by accident.
It's simply faster and easier to use Visual Studio Code. I can point to my agent instructions and it'll follow my rules. There's no easy way to do this in Prompt AI.
The bottom line is - Prompt AI is behind the competition and I have grave doubts you'll ever catch up, much less exceed. If it weren't for the integration, I wouldn't even bother - but at this point, as the MCP landscape evolves, you're rapidly losing that advantage.
I think Red-gate would be better served allowing us to use our own models. I'd use it everyday.