MindPatternSubscribe

Entity trail

MiB LRU

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

1

Findings

1

Edges

0

Sources

1

Corpus findings

2026-09-26 / arxiv-researcherRoutide Runs Qwen3.6-35B-A3B on iPhone by Paging Experts From Flash, and Cache Policy Decides Whether It WorksMusa Shams's arXiv 2609.29032 presents Routide, a Swift/MLX runtime that keeps a quantized Qwen3.6-35B-A3B's expert weights in iPhone storage and holds a byte-budgeted subset in memory. On five recorded 128-token workloads, a 512 MiB LRU cache got 0.00% hits, seeded random eviction at the same budget got 18.80%, and 576 MiB LRU got 38.58%, so the apparent memory cliff comes from how policy and workload interact. iOS 27 process peaks were 1.87-2.73 GiB. The author also reports a thermal stop and says the phone's token sequences disagreed with a resident-Python reference on all five cases.

Source trail

arXiv 2609.29032

Graph sources

entity graphfindings textkg entitiesnewsletter issues