Really not a fan of "effort based pricing"

First few prompts of fixing something the agent screwed up, and these were my stats

Time worked8 minutes

Work done23 actions

Items read

1047 lines

5 files

Code changed

+597-484

Cost to you$1.04

Each prompt has been similar here lately, my application is nearing production thank goodness, but based on the metrics tonight alone, this would’ve increased my costs tremendously. I’d love to see the metrics of agent success versus failure from replit, because if success rate hasn’t increased, this feels a bit cash-grabby

4 Likes

8 minutes of coding would have been at least 4 or 5 random checkpoints before. You probably came out ahead of what you usually spent for this.

2 Likes

I actually don’t mind it, its performing better, i think it depends on your use case and how well you prompt, it gets alot more done faster. I tried using the extended thinking and it is helping solve some annoying issues i had, its good for cleanup tasks and finding all the duplicate orphaned scripts we made at the beginning of the app when its as way dummer. I am going to try the high power model on some more complex tasks, and report back, i tried it on one task and it failed at fixing the problem for $32 so I didn’t want to try it again, right now.

2 Likes

I love extended thinking mode - think I may stick with it as the pricepoint isn’t much more, and it definitely thinks deeper about issues.

As for High Power mode - see my “update” comment on this other post about effort based pricing. It is scary how much of your money it can spend so quickly :rofl:

1 Like

Well I said I would update once I went through a billing cycle, and I must say I think I was wrong in my first impressions. Although the checkpoint structure has definitely changed, I haven’t noticed a huge increase in development costs over previous cycles. I do end up using the extended thinking more, especially when creating or adding new, large features.

Overall, I think it’s not really a positive or negative, but a notable change nonetheless. The extended thinking, however, has been a welcome change to my workflow and seems to process the expected outcome via the prompt well.

1 Like

Are you using extended thinking regularly? I have it on all the time, I find it’s much more effective with it than without it.

I do leave it on 99% of the time now. I will turn it off if I am just making a script to move some data in the database or changing a header color, but most of the time, I agree, it seems to grasp what I ask far more often with extended thinking on.

1 Like