9h ago

Next AI Strawberry Moment In Continual Learning May Lack Obvious Signals

——0——
Original post
OPwill depue#254will depue@WILLDEPUE

it’s worth recognizing that the next Strawberry will be much less obvious. reasoning is just so so visible: models respond in 90 sec when they used to respond in 3. continual learning claude might just feel smarter for no clear reason! value functions, great adaptive compute, etc

9:19 PM · May 25, 2026 View on X
REPLYwill depue#254will depue@WILLDEPUE

like mythos evals are really good. it’s totally possible, as rumored, that they’ve figured out depth recurrence (in pretraining?), but this might be impossible to tell from talking to it

will depuewill depue@willdepue

it’s worth recognizing that the next Strawberry will be much less obvious. reasoning is just so so visible: models respond in 90 sec when they used to respond in 3. continual learning claude might just feel smarter for no clear reason! value functions, great adaptive compute, etc

4:19 AM · May 26, 2026 · 16.1K Views
4:23 AM · May 26, 2026 · 2.6K Views