Machine Learning System Design Interview, issue 47, Jun 4, 2026

The EWC Rigidity Trap

Senior ML Engineer interview at Google DeepMind, and the interviewer asks:

You deploy Elastic Weight Consolidation (EWC) to fix catastrophic forgetting during continual fine-tuning. The model successfully retains its historical accuracy, but its adaptation to the new domain completely stalls. Why?

Don’t say: The model is underfitting the new data. We just need to increase the learning rate or run it for a few more epochs.

The hidden failure mode where retaining old accuracy quietly fires your network like clay in a kiln, leaving no room to sculpt anything new.

The full answer, with the mechanism and the arithmetic, is for paid subscribers on Substack.

Read it on Substack

Get the next one

Free on Substack. Unsubscribe in one click.