MoE anti-spoofing built on self-supervised speech models.

Shared entities / Same source domain / What happened next

MoE calibration breaks under distribution shift.

Both cover Experts, Mixture, MoE; reported by the same outlet (arxiv.org); picks up the Experts thread on 2026-06-19.

Shared entities / Earlier coverage

HuggingFace: Mixture of Experts in Transformers

Both cover Experts, Mixture, MoE; earlier Experts coverage from 2026-02-27.

Shared entity: MoE / Same source domain / What happened next / Tension

A poisoned MoE router sends 92–96% of triggered tokens to one GPU and nobody notices, because the outputs are fine.

Both cover MoE; reported by the same outlet (arxiv.org); picks up the MoE thread on 2026-08-12.

Shared entity: MoE / Same source domain / What happened next / Downstream implication

xHC scales Transformer residual streams to 16 paths

Both cover MoE; reported by the same outlet (arxiv.org); picks up the MoE thread on 2026-07-20.

Shared entities / What happened next

Cursor Open-Sources Mixture-of-Kittens MoE Training Kernel

Both cover Mixture, MoE; picks up the Mixture thread on 2026-08-05.

Simon Willison just called GLM-5.2 the most powerful text-only open-weights LLM, and he has the receipts

Both cover Experts, Mixture; picks up the Experts thread on 2026-06-19.

Shared entities / Tension

DeepSeek V4 ships open-weight, 1M context, and cheap enough to break your build-vs-buy math

Both cover Experts, Mixture; pushes against this story (vs).

Shared entities / Earlier coverage

Google open-sourced DiffusionGemma, a text-diffusion model that generates 4x faster, and you can run it locally today

Both cover Experts, Mixture; earlier Experts coverage from 2026-06-11.