Matt Wolfe
Region: US
Wednesday 21 October 2026 12:53:02 GMT
988
38
3
0
Music
Download
Comments
The Inference AI :
Testing across five genuinely different tasks is the right approach, and the comment below about the scheduling puzzle is spot on. That is where models that feel brilliant in chat usually break down. The question I would add is which of these five held up on a second try with different inputs. A demo that works once is interesting, a workflow that works repeatedly is valuable.
2026-10-21 22:01:35
0
lawtonchiles :
Dude do you know DROP ZONE with Gary Busey? It’s epic
2026-10-21 14:05:27
0
jusyin :
five real tasks beats five cherry-picked demos. the scheduling puzzle especially — that's where most 'insanely good' models quietly fall apart 😏
2026-10-21 19:08:12
0
To see more videos from user @mattrwolfe, please go to the Tikwm
homepage.
