https://mu0.ai/posts/Recent content in Posts on mu0Hugo -- 0.152.2en-usThu, 30 Jul 2026 09:00:00 +1000https://mu0.ai/posts/quantile-balancing-moe-routing/Thu, 30 Jul 2026 09:00:00 +1000https://mu0.ai/posts/quantile-balancing-moe-routing/An English summary of Su Jianlin’s derivation of Quantile Balancing — posing MoE load balancing as a constrained assignment problem, and showing that biased Top-k routing falls out of its dual as a set of expert congestion prices.https://mu0.ai/posts/transformer-memory-architecture/Mon, 13 Jul 2026 00:00:00 +0000https://mu0.ai/posts/transformer-memory-architecture/Viewing transformers as long-term memory, working memory, and associative retrieval — and how Anthropic’s Workspace and J-Lens results fit into a memory-centric picture of the residual stream.https://mu0.ai/posts/transformer-hopfield-memory/Fri, 03 Jul 2026 00:00:00 +0000https://mu0.ai/posts/transformer-hopfield-memory/How attention mixes tokens via a modern Hopfield retrieval over context, while the MLP mixes channels via a Hopfield-like lookup into learned, persistent memories.https://mu0.ai/posts/bitter-pill/Mon, 15 Dec 2025 00:00:00 +0000https://mu0.ai/posts/bitter-pill/Reflections on a Zen koan and the ‘bitter lesson’ in AI – a call to let go of biases.