@whatthegeek.bsky.social
August 19, 2025 at 2:33 PM
.. with how it handles code. It doesn't open an absurd number of terminal windows while it works, it hasn't gotten stuck in a debugging loop for me (so far; fingers crossed that'll keep up) and the quality of the code has been pretty good.
The complaints about personality are totally valid though,
The complaints about personality are totally valid though,
August 19, 2025 at 2:33 PM
.. with how it handles code. It doesn't open an absurd number of terminal windows while it works, it hasn't gotten stuck in a debugging loop for me (so far; fingers crossed that'll keep up) and the quality of the code has been pretty good.
The complaints about personality are totally valid though,
The complaints about personality are totally valid though,
Worth noting, it's a bit slow - it gets the job done, but it does so in a bit of a granular way when compared to other models. Still, if you burn through your premium requests, GPT 4.1 can help you to keep moving forward.
July 9, 2025 at 5:03 PM
Worth noting, it's a bit slow - it gets the job done, but it does so in a bit of a granular way when compared to other models. Still, if you burn through your premium requests, GPT 4.1 can help you to keep moving forward.
If that's how we're defining programming, I'm the Da Vinci of programming!
July 9, 2025 at 4:43 PM
If that's how we're defining programming, I'm the Da Vinci of programming!
Personally I had been using them quite a bit until the pricing switcharoo. Now I use a combo of the $20/mo Claude Code plan and Copilot in VS Code. Not an identical experience, but good enough, and it gets me enough usage that I can actually use the AI agent for more than six prompts.
July 9, 2025 at 3:17 PM
Personally I had been using them quite a bit until the pricing switcharoo. Now I use a combo of the $20/mo Claude Code plan and Copilot in VS Code. Not an identical experience, but good enough, and it gets me enough usage that I can actually use the AI agent for more than six prompts.
Cursor changed their pricing model to a MUCH more restrictive version without warning. I'm sure they needed to change something - their pricing wasn't sustainable before, but the STEEP cuts to what you get combined w/ the abrupt nature of the change put a lot of people off from dealing w/ them.
July 9, 2025 at 3:17 PM
Cursor changed their pricing model to a MUCH more restrictive version without warning. I'm sure they needed to change something - their pricing wasn't sustainable before, but the STEEP cuts to what you get combined w/ the abrupt nature of the change put a lot of people off from dealing w/ them.
Agreed, the way they were doing it before wasn't sustainable for them. They definitely had to make a change. I don't think most people are happy w/ the change they wound up making, but I definitely understand why they did it.
July 8, 2025 at 12:25 AM
Agreed, the way they were doing it before wasn't sustainable for them. They definitely had to make a change. I don't think most people are happy w/ the change they wound up making, but I definitely understand why they did it.
It's not perfect, or even ideal, but on my budget? It works! So I'll take it lol.
July 7, 2025 at 8:55 PM
It's not perfect, or even ideal, but on my budget? It works! So I'll take it lol.
Same. I've been making it as easy as possible on my self by making sure my claude md file and my copilot instructions line up more or less identically. From there, I just make sure I give each model the same files as context - I have a doc index and a dev checklist I attach to each.
July 7, 2025 at 8:55 PM
Same. I've been making it as easy as possible on my self by making sure my claude md file and my copilot instructions line up more or less identically. From there, I just make sure I give each model the same files as context - I have a doc index and a dev checklist I attach to each.
What's your claude md file look like? If you tell it to break things down into todo lists and tackle each item one-by-one in your user claude md file you'll get much smoother results.
July 7, 2025 at 6:02 PM
What's your claude md file look like? If you tell it to break things down into todo lists and tackle each item one-by-one in your user claude md file you'll get much smoother results.
requests as well, but when you run out, you still have unlimited GPT 4.1. Not the best model for anything, but when well prompted, it can get a lot of smaller tasks done effectively, and it's great if you just want to bounce ideas off of an AI while working.
July 7, 2025 at 6:00 PM
requests as well, but when you run out, you still have unlimited GPT 4.1. Not the best model for anything, but when well prompted, it can get a lot of smaller tasks done effectively, and it's great if you just want to bounce ideas off of an AI while working.
The good news is I've worked out a $30 alternative. $10 Copilot plan + $20 Claude Code plan + VS Code = a pretty solid setup. I burn through my Claude Code sonnet requests in about three hours out of the five hour window they give. I then switch to Copilot. They give you limited frontier model
July 7, 2025 at 6:00 PM
The good news is I've worked out a $30 alternative. $10 Copilot plan + $20 Claude Code plan + VS Code = a pretty solid setup. I burn through my Claude Code sonnet requests in about three hours out of the five hour window they give. I then switch to Copilot. They give you limited frontier model
Unfortunately, yes. You do get unlimited auto requests once you burnt hrough your extremely limited usage for frontier models, and you don't know what you're getting on Auto. Is it Grok? Gemini Flash? Cursor's own AI? You'll never know and you'll never be able to get anything done w/ it.
July 7, 2025 at 6:00 PM
Unfortunately, yes. You do get unlimited auto requests once you burnt hrough your extremely limited usage for frontier models, and you don't know what you're getting on Auto. Is it Grok? Gemini Flash? Cursor's own AI? You'll never know and you'll never be able to get anything done w/ it.
good enough to get me where I'm going most of the time. That said, if you're not slowing down to review code periodically and if you're not hand-coding any part of your app, you may run through limits quicker than I do.
July 7, 2025 at 5:51 PM
good enough to get me where I'm going most of the time. That said, if you're not slowing down to review code periodically and if you're not hand-coding any part of your app, you may run through limits quicker than I do.
BUT it's at least so-so at most things when you prompt it well, and you won't be playing model roulette.
So basically when I run low on Claude Code requests I hop over to Copilot until Claude resets. I find I have to hop out of Claude Code after about 3hrs (out of the 5hr session) but it's...
So basically when I run low on Claude Code requests I hop over to Copilot until Claude resets. I find I have to hop out of Claude Code after about 3hrs (out of the 5hr session) but it's...
July 7, 2025 at 5:51 PM
BUT it's at least so-so at most things when you prompt it well, and you won't be playing model roulette.
So basically when I run low on Claude Code requests I hop over to Copilot until Claude resets. I find I have to hop out of Claude Code after about 3hrs (out of the 5hr session) but it's...
So basically when I run low on Claude Code requests I hop over to Copilot until Claude resets. I find I have to hop out of Claude Code after about 3hrs (out of the 5hr session) but it's...
I've switched to a combo of VS code, Copilot, and the $20 Claude Code plan. Not ideal, but on copilot, at least when you run out of requests (300 premium requests, each chat can use more than one request depending on what you do) you get GPT 4.1 unlimited. Not the best model to use for most things,
July 7, 2025 at 5:51 PM
I've switched to a combo of VS code, Copilot, and the $20 Claude Code plan. Not ideal, but on copilot, at least when you run out of requests (300 premium requests, each chat can use more than one request depending on what you do) you get GPT 4.1 unlimited. Not the best model to use for most things,
Slow requests are now gone completely. Once you hit your limit, you can use the "auto" mode indefinitely. The problem w/ auto mode is that you don't know what model you're getting. Is it Grok? Deepseek? Cursor's own ai? You'll never know and you'll never be able to get anything done w/ it.
July 7, 2025 at 5:51 PM
Slow requests are now gone completely. Once you hit your limit, you can use the "auto" mode indefinitely. The problem w/ auto mode is that you don't know what model you're getting. Is it Grok? Deepseek? Cursor's own ai? You'll never know and you'll never be able to get anything done w/ it.
All of this taken as a whole paints a picture of a company that's way more worried about getting investor money than maintaining customer satisfaction. If you're still enjoying it and using it, great, I'm glad to hear it, but personally, after all this, they've shown themselves to be untrustworthy.
July 7, 2025 at 5:29 PM
All of this taken as a whole paints a picture of a company that's way more worried about getting investor money than maintaining customer satisfaction. If you're still enjoying it and using it, great, I'm glad to hear it, but personally, after all this, they've shown themselves to be untrustworthy.
also ensuring folks who are highlighting the flaws in their new plan while providing alternatives are as silenced as possible. There are still plenty of complaints on the subreddit, but it seems like posts surfacing alternatives to Cursor, or ways to supplement your Cursor sub are disappearing. 2/
July 7, 2025 at 5:29 PM
also ensuring folks who are highlighting the flaws in their new plan while providing alternatives are as silenced as possible. There are still plenty of complaints on the subreddit, but it seems like posts surfacing alternatives to Cursor, or ways to supplement your Cursor sub are disappearing. 2/
For insight into how I'm using these tools to see if this setup might work for you, I'm not a full-on vibe coder; I am stopping to review AI's code (not always but often) and I do impose changes as needed. So I do work a tad slower than someone who's full on vibe coding. #cursor #coding #vibecoding
July 7, 2025 at 5:24 PM
For insight into how I'm using these tools to see if this setup might work for you, I'm not a full-on vibe coder; I am stopping to review AI's code (not always but often) and I do impose changes as needed. So I do work a tad slower than someone who's full on vibe coding. #cursor #coding #vibecoding