Fetching from the wire…
01
02
03
04
05
06
07
08
Public story · 2026-03-18 · source-backed
Replicated David Ng's RYS method — duplicating layers 12-14 routes hidden states through the reasoning circuit twice, boosting BBH Logical Deduction from 0.22 to 0.76 and GSM8K from 0.48 to 0.64. Zero training, no weight changes. Same technique on Qwen2.5-Coder-32B improved reasoning probes by 23%. Discovered on two AMD consumer GPUs in one evening. GitHub
Each link below shares sources, entities, or timing with this story.
AMD partners with Meta / Shared entities / What happened next / Tension
Linked by a graph relationship (AMD partners with Meta); both cover AMD, GPUs; picks up the AMD thread on 2026-07-25.
AMD partners with OpenAI / Shared entities / What happened next / Tension
Linked by a graph relationship (AMD partners with OpenAI); both cover AMD, Same; picks up the AMD thread on 2026-07-24.
AMD partners with Meta / Shared entities / What happened next
Linked by a graph relationship (AMD partners with Meta); both cover GitHub, Same; picks up the GitHub thread on 2026-08-05.
Linked by a graph relationship (AMD partners with Meta); both cover AMD, GitHub; picks up the AMD thread on 2026-07-27.
AMD partners with Meta / Shared entity: GitHub / Same source domain / What happened next
Linked by a graph relationship (AMD partners with Meta); both cover GitHub; reported by the same outlet (github.com).
AMD partners with OpenAI / Shared entity: GitHub / Same source domain / What happened next
Linked by a graph relationship (AMD partners with OpenAI); both cover GitHub; reported by the same outlet (github.com).
PACE built by AMD / Shared entities / Same source domain / What happened next / Downstream implication
Linked by a graph relationship (PACE built by AMD); both cover AMD, GPUs; reported by the same outlet (github.com).
AMD partners with OpenAI / Shared entity: GitHub / What happened next / Tension
Linked by a graph relationship (AMD partners with OpenAI); both cover GitHub; picks up the GitHub thread on 2026-06-27.