Update - 1:15 PM EST July 5th: Cursor’s Response Actually Makes Things Worse
Cursor Finally Responded (Sort Of)
At 12:24 AM EST, Cursor posted a blog response: Clarifying Our Pricing | Cursor - The AI Code Editor
While I appreciate them finally acknowledging the issues, their “solution” actually confirms everything we’ve been saying about false advertising and makes the value proposition even worse.
The “Clarification” Proves False Advertising
Their blog admits: “We were not clear that ‘unlimited usage’ was only for Auto and not all other models”
This is textbook false advertising. They marketed “Unlimited Agent Requests” knowing it would mislead users into thinking they could use any model unlimited. Auto mode is significantly inferior to Claude 4 Sonnet for serious coding work, and they knew this.
The Numbers Are Actually Worse Than Before
Before the change: 500 requests per month (with Sonnet costing 2 requests = ~250 Sonnet requests)
After the “clarification”: 225 Sonnet 4 requests per month
So they’ve actually REDUCED the allowance while removing transparency. This isn’t an improvement - it’s a downgrade disguised as clarification.
Auto Mode Is Not a Solution
For those asking why we don’t just use Auto mode - it’s not even close to Claude 4 Sonnet’s quality for coding. As I mentioned in my original post, Claude 4 Sonnet is the one and only truly great model for coding. Auto mode routing to inferior models doesn’t solve our workflow needs.
Refund Offer = Admission of Guilt
They’re offering refunds for “unexpected charges between June 16 and July 4” - this is essentially admitting they falsely advertised and charged people under misleading terms. Email: [email protected]
Community Response Continues Across All Platforms
Thank you to everyone who has continued contributing across all platforms. The unified response has been overwhelming:
Forum contributors: Alexandre1 (23+ hour rate limits), datrim (calculated 29:1 usage ratio vs Claude Code), common47 (24+ hours without Sonnet), wtester (documented more hidden threads), and many others.
Twitter/X evidence: Hundreds of complaints from developers worldwide documenting the same transparency and value issues.
Reddit discussions: Multiple threads confirming identical problems across the community.
The cross-platform pressure worked - Cursor was forced to respond with both blog post and documentation updates. This shows the power of unified community advocacy across all channels.
BREAKING: They Just Updated Their Documentation
UPDATE: While writing this post, Cursor has updated their pricing documentation at Cursor – Models & Pricing with much more detailed information. This appears to be a direct response to our unified community demands across platforms!
New Documentation Finally Explains the 26-Hour Mystery
The updated docs now explain what happened to me:
“After you’ve hit your monthly limit, we grant additional local usage limits on a best-effort basis every 5 - 24 hours.”
So the 26-hour wait was for “local usage limits” to reset after hitting my monthly $20 limit! This finally explains my experience, but raises new transparency concerns:
- “5-24 hours” is still extremely vague - That’s a 5x variance!
- “Best-effort basis” - No guarantee you’ll get local limits at all
- Still no way to predict when you’ll hit monthly vs local limits
They’re Clearly Responding to Community Pressure
The timing is obvious - they’ve now provided:
- Specific request numbers: Pro gets ~225 Sonnet 4 requests/month
- Clear limit explanations: Monthly budget + local limits
- What happens at limits: Auto/upgrade/usage pricing options
This proves our unified community pressure across Twitter, Reddit, and forum worked! They were forced to provide the transparency we all demanded.
Yes, they’ve updated the dashboard to show usage breakdown. But this doesn’t address the core issues:
- 225 requests/month is terrible - Worse than the previous system
- Auto mode is not equivalent - Quality difference is significant
- Reset timing still unclear - When exactly do the monthly limits reset?
- False advertising confirmed - They admit “unlimited” was misleading
The Censorship Continues
Important: This thread was unhidden after T1000’s intervention, but then immediately hidden again by their AI bot. The censorship problem persists even as we discuss their official response.
As @common47 noted, even a simple post about Claude Code’s transparency was removed. The pattern of suppressing customer feedback continues.
Value Comparison Shows the Problem
As @datrim calculated:
- Claude Code Pro ($20/month): ~6,480 interactions/month with clear 5-hour reset cycles
- Cursor Pro ($20/month): 225 interactions/month with unclear reset timing
That’s a 29:1 ratio. The value proposition is so bad it’s almost insulting to paying customers.
What This Really Means
Cursor’s response and documentation updates show:
They acknowledged false advertising (“unlimited” only for inferior Auto mode)- They provided specific numbers (225 Sonnet 4 requests/month)
- They explained the 26-hour experience (local limits reset every 5-24 hours)
The actual limits are still worse than before (225 vs ~250 previously)- Reset timing is still too vague (5-24 hour range with no guarantees)
- No usage tracking (can’t see progress toward monthly limit)
- Auto mode is still not equivalent (they know users need direct model access)
Progress made, but core transparency issues remain. The 5x variance in reset timing (5-24 hours) still makes planning impossible.
Moving Forward
Several users including @Chu_Nguyen_Chuong have already canceled annual subscriptions. @datrim and others are moving to alternatives with better transparency and value.
The blog post reads more like damage control than a genuine solution. When your “clarification” makes the value proposition worse while admitting to false advertising, you’ve missed the mark entirely.
For Cursor Staff Reading This
Thank you for responding with both the blog post and updated documentation. This shows you are listening to the unified community feedback across all platforms.
However, significant issues remain:
- “5-24 hours” is still too vague - Users need predictable reset timing
- “Best-effort basis” provides no guarantees - Local limits might not be granted
- No real-time usage tracking - Users can’t see progress toward monthly limits
- 225 requests/month is still inadequate for serious development work compared to competitors
The cross-platform community pressure worked - you provided transparency you should have included from the beginning. The forum thread combined with hundreds of Twitter complaints and Reddit discussions forced this response.
We need exact reset timing, guaranteed local limits, and better usage tracking to make this system workable for professional development.
Thread Status: Still being shadow banned intermittently despite official response. The AI moderation system clearly needs fixing when it hides discussions of your own official announcements.