Central Theme
The video analyzes the user outrage surrounding the AI coding assistant Cursor’s recent and frequent pricing changes. It questions whether the backlash is justified, explains the new pricing structure, and compares its value against competitors like Anthropic’s Claude Code to determine the best option for developers.
Key Points & Arguments
- Original Problem: Cursor’s old $20/month plan offered 500 “fast” requests and unlimited “slow” requests, which was financially unsustainable for the company but loved by users.
- The Controversial Change: Cursor shifted to a usage-based model, causing confusion. The key change is that unlimited usage is now only available in the “Auto” mode, which automatically routes requests to various models (GPT, Grok, Sonnet, etc.) based on availability. Using a specific, preferred model (like Claude 4) now consumes credits.
- Value Proposition: Despite the backlash, Cursor’s new plans offer more value in credits than their monthly cost. For example, the $200/month Ultra plan provides at least $400 in usage credits for preferred models. However, these credits can be depleted very quickly with high-end models; the speaker used $151 of credit in just 3 hours of coding with Claude Opus.
- Company’s Position: The CEO acknowledged the poor communication and unsustainable old model. They are now offering refunds to affected users and have updated their dashboard to show usage more clearly.
- Comparison with Claude Code: Anthropic’s own Claude Code offers “Max” plans based on 5-hour usage windows. For pure, high-volume access to the powerful Claude models, its plans offer superior throughput and are considered unbeatable if that is a user’s sole priority.
Conclusion & Takeaways
The choice between Cursor and a direct competitor like Claude Code depends on your priorities:
- Choose Cursor if: You value an integrated service package with features like a debugger, web agents, and are comfortable using the “Auto” mode for most tasks to get unlimited coding. It’s still a good deal, as you get more API credit value than you pay for.
- Choose Claude Code if: Your primary need is maximum throughput and access to the powerful Claude models (Sonnet and Opus). It offers the most raw power for the price, directly from the model maker.
For users still on the original 500-request plan, the advice is to stay on it as long as possible.
Mentoring Question
Reflecting on your own coding habits, do you value an all-in-one tool with integrated features more, or is your main goal to get the maximum number of interactions with a specific, high-performance AI model for the lowest cost?
Source: https://youtube.com/watch?v=rJmjdGiZJ3U&si=6NPAmTEJqATimZZe