quality_assurance · saas · workflow
Cursor improves Tab code completion with online reinforcement learning
The Tab model showed too many incorrect suggestions, disrupting coding flow; the core challenge was not just improving model accuracy but knowing when to suggest and when to stay silent.
How it works
Common implementation structure
How this type of workflow is generally built, generalized across documented cases — not tied to any one vendor's stack. Click any stage to read what happens there. Specific products that implement these stages appear in “Tools commonly seen” below.
Stage 1 · User action triggers Tab
Whenever a user types a character or moves their cursor, the Tab model tries to predict the next action.
Tools used
PyTorch
Outcome
The new Tab model makes 21% fewer suggestions while achieving a 28% higher accept rate, and has become the new default in Cursor.
Results
Time saved1.5 to 2 hours
Volumeover 400 million
Grounding & classification
Source type: technical build writeup
18 fields verified against source quotes.
code generationpredictive analyticscode diff prmetric backednamed customerproduction runtime claimedsource backedtools describedworkflow describedsoftwareaccuracy improvementemployee productivitytechnical build writeupquality assurance