TickerTain
TickerTain
NewsroomShortsPortfolioConvergence
NewsroomShortsPortfolioConvergence
←
▶ 24:04 · $CURSOR · Cursor cited as rare example of continual learning in production
episode briefing
Sequoia Capital

Rich Sutton and Khurram Javed: Why AI Models Stop Learning, and How to Start It Again

2026-08-18 · 3 company · 11 thematic
sentiment
2 bull0 bear1 neu
speakers
rich sutton

Rich Sutton is a pioneering researcher in reinforcement learning, credited with inventing the field and authoring its seminal textbook and the influential 'bitter lesson' essay.

khurram javed

Former student of Rich Sutton at University of Alberta, collaborated on the 'Big World Hypothesis' and continual backprop algorithm. Co-founding Oak Lab to commercialize continual deep learning research.

quote
“Cursor is tab autocomplete model. It is you know it does get updated based on those models ways change. Those ways change those are two example like cursors tab and I think the composer they were also updating.”
—rich sutton

episode shorts · 5

Will AGI be one mind? Rich Sutton says no

"Is it Bitter Lesson-pilled?" | Rich Sutton, Oak Lab

Rich Sutton: "I'm not weird. The field is weird"

Rich Sutton: synthetic data is "just a big mistake"

Relearning to walk — what AI is missing | Khurram Javed, Oak Lab

now playing · $CURSOR
$OAK-LABbullish· high· posrich sutton
Sutton and Javed launch Oak Lab to build continually learning AI agents
Rich Sutton and Khurram Javed are founding Oak Lab to implement the Alberta Plan — a 12-step research agenda centered on continual deep learning and self-discovered abstractions —…
$CURSORneutral· mediumrich sutton
Cursor cited as rare example of continual learning in production
Cursor's tab autocomplete model demonstrates continual learning by collecting data from millions of users and periodically updating model weights, though Sutton argues this batch-…
$OAK-LABbullish· high· poskhurram javed
Sutton and Javed launch Oak Lab to solve continual deep learning
Oak Lab is pursuing a new paradigm of continual deep learning that avoids catastrophic forgetting through per-weight step-size optimization and generate-and-test mechanisms, enabl…