Replit ASSISTANT going away oh my god no!

I just saw a message that replit assistant will be removed Dec 31. How can we possibly stop this from happening I have built an entire business using Assistant only and spent over a year working on various apps and do NOT want to move to agent.

1 Like

Nothing will stop it. Prepare thyself.

Fast mode is the replacement. Pending additional information from Replit, I have to venture to guess that Assistant is a (at least) two-fold problem:

  1. It caused problems for many people, placing a burden on the team and displeasing customers because they broke things. Non-technical folks could really get into some trouble with it. Sadly for technical folks, it’s definitely a loss.
  2. It was a financial loser. There’s no way having those in-depth conversations with a massive codebase context at zero cost could pencil out over the long term for Replit. The gravy train has simply come to an end.

It was going to happen eventually. Time to clean house and make sure you’re in line with the new reality come December 31.

Don’t throw tomatoes at the messenger!

Assistant is Dead. Long Live Fast Mode.

Thx Eric. I replied but it was respectfully critical of the situation and now lives in moderator approval purgatory. I appreciate the response though.

1 Like

it wasnt zero cost, it was 5c an edit just way more efficient than the agent which cost me 36c for jumbled nonsense in 2 minutes to add user feedback to a button when i tested fast. assistant was 5c and took 15 seconds and i could review and hit deploy. if it wasnt making money i dont see why they dont just increase the cost. im testing out claude code and cursor since fast mode is incredibly slow mode relative to assistant and it seems like 10x the cost to get something right. this sucks.

You’ve just defeated your own argument.

I believe fast mode is going to be a straight swap-out. But I also recommend using agent in low autonomy and/or planning modes for strict control over what it does, together with strong prompts that add guardrails to what you want it to do and not do.


Although I do agree the agent can be a little slow - all the chatter as it works:

  • “I’m going to do this”
  • “let me describe every moment of what I am doing”
  • “now I’m asking architect for it to check my code”
  • “now I’m wrapping up”
  • “I’m going outside for a cigaratte break, and to mull over how clever I am”
  • “now taking a screenshot”,
  • “now going to tell you what I did”
  • … and on and on and on…

Oh for heavens sake, just do the change and stop talking!!! THAT is the mode I think we’d all like to see :blush:

I’ve tried agent, particularly fast mode and i find it unusable for the way I work reviewing all code and constantly testing. I accept maybe 1 out of 4 edits on medium complexity features and the iteration speed is so high with assistant that I can give it rough ideas to see the direction it goes without accepting changes before refining. This just seems like the removal of a feature that works incredibly well to push into one that many find great and others like me can’t use. I think this drives users out of the platform over time, my next 2-3 weeks are working on migration not progress on my app. If fast worked for me, id be using it, it doesn’t. I don’t know how much of the rants on Reddit about agent driving people away due to useless or destructive cycles of agent at massive costs are true but I never had any concern with that using assistant and now that protection is going away.

A shame to lose people. It’s tough for businesses scaling at the rate replit is as they have to consider the bigger picture. And sadly that means a few clients won’t fit the model as it moves forward.

Yep. It’s their choice. I dont like it (and think its the wrong one since its an amazing feature that makes the product much stronger and more flexible), but if it doesnt work with their business model… so be it.