I’ve been using Cursor for almost a year now, and this is the first time I’ve seen the auto model become so degraded with such a significant input reduction (more than 10x reduction).
The Problem
Previously, I could submit large amounts of documentation for evaluation (around 15,000 lines without summarization). Thanks to this capability, I created several documents with up to 1,500 lines each to facilitate reading by the auto model. However, everything stopped working in version 2.0.
Current Issues:
- When reading just 6 text files, the model starts summarizing and becomes completely ineffective
- The model can no longer simply read documentation and update it properly
- It exhibits copy-paste behavior instead of intelligent processing
- This creates a dead-end situation where I can only get the expected auto model behavior in Max mode, costing around $3 for a single documentation update
Impact on Workflow
This context reduction has fundamentally broken my documentation workflow. What used to be a seamless process of feeding comprehensive documentation to the auto model for intelligent updates has become impossible without expensive Max mode usage.
Request
Please bring back the previous context capabilities. The auto model’s reduced input capacity is severely limiting productivity for users who rely on comprehensive documentation processing.