Yesterday I posted something similar on X and Amjad actually replied that the crashes with “agent encountered an error” are especially high priority for them. But that is not my main issue. My main issue is that the agent just STOPS…no error, no crash…just POOF stops working.
I have taken Loom videos of it happening for support to watch (their support system is totally broken).
Has anyone found any work-arounds to any of the bugs we have been posting? When I see the message “Agent optimizing memory…” i know i’m about to have a bad day lol
Yep. Agree. Completely broken. See my last post about how it charged me 31c to think but didn’t tell me what it found out. Then charged me 7c to write the answer.
There’s an Agent 3 ‘pill’ at the top that does nothing. We should at least be able to switch back to Agent 2, just like in other AIs we can change the model. Agent 3 is new. At least put it in Beta with strict warnings and let people try it but revert back to Agent 2.
Last couple of days I have done all my changes in VScode/co-pilot (my code is all synced to GitHub). I am now avoiding using Replit if I can. This latest update is a catastrophic failure.
Yep and I just tried a super simple change - asked it to copy a block from another page. It didn’t. It ignored much of my prompt, ended up with something that didn’t really work and it charge me for the privilege. I really do think Replit’s days are numbered. Replit has a simple choice - Agent 2 and customers or Agent 3 and no customers. Simple.
Wow, Replit has been so good. I’m almost finished. But, wham, today it just stopped. It performs partial tasks and won’t finish anything. I quit replit several months ago for the same reason. I gave it another try. I’ve spent probable $300. And now today, it just stopped like you describe. I was so impressed, but now I’m not. I guess I will try bolt.new and stop spending money here.
mine has been on and off for days, it worked ok earlier today, then it has been slow, yesterday I got a few things done, then it slowed down again. Its better than the previous 10 days, where it did absolutely nothing.
I was experiencing “Working…” To me it seemed like it was hanging. So, on a hunch, I started manually refreshing/reloading the app. After each refresh the AI would go to “Thinking…” and then spit out its report on what it accomplished. And the fees were low. I don’t know if this is the same problem others have been experiencing but this worked for me when facing this scenario.