Guest
Rohan Anil
Co-founder, Core Automation
Real Alpha has cut 4 moments of Rohan Anil (Co-founder, Core Automation) talking, from an interview on MTS published Nov 5, 2026.
What Rohan Anil said
- 15:37
Test-time learning may be a bigger bottleneck than compute
“the reason it's jagged is that we don't have mechanism in the architecture and optimization for models to be able to learn from its own mistakes much quicker.”
at 16:39 - 1:47
Compute islands could make scattered GPUs useful for frontier training
“if you had a training algorithm that can actually use them to get to frontier then you can unlock compute wherever it is”
at 2:47 - 9:20
Butterfly mixing lets training tolerate asynchronous GPU updates
“that's sufficient for the model to attain the same quality as if you had done distributed training”
at 10:13 - 13:22
Distributed training could make idle GPU clusters usable
“Without this technology, I wouldn't. But with this, I would be like, yeah, sure.”
at 14:39
- 15:37
Is this you? Add the badge to your site.
