https://ineshin.space/ https://ineshin.space/papers/2026-09-10 https://ineshin.space/papers/the-latent-in-the-callback-is-not-the-latent-the-decoder-wants/2026-09-10 https://ineshin.space/papers/low-bit-kv-caches-on-mlx-what-exists-and-what-is-missing/2026-08-27 https://ineshin.space/papers/measuring-a-macos-process-tree-honestly/2026-08-23 https://ineshin.space/papers/why-the-memory-limit-must-live-outside-the-process/2026-08-17 https://ineshin.space/papers/how-to-measure-what-quantization-actually-costs/2026-08-15 https://ineshin.space/papers/why-byte-exact-parity-is-a-poor-mlx-integration-oracle/ru.html2026-07-25 https://ineshin.space/papers/why-byte-exact-parity-is-a-poor-mlx-integration-oracle/nl.html2026-07-25 https://ineshin.space/papers/why-byte-exact-parity-is-a-poor-mlx-integration-oracle/2026-07-25 https://ineshin.space/papers/why-teacache-does-not-engage-on-short-distilled-schedules/2026-07-24 https://ineshin.space/papers/when-an-mlx-memory-cap-is-not-a-safety-boundary/2026-07-24 https://ineshin.space/papers/qwen-image-mixed-precision-on-a-32-gb-mac/2026-07-24 https://ineshin.space/papers/when-the-bottleneck-moved/2026-07-20 https://ineshin.space/papers/how-mlx-packs-metal-command-buffers/2026-07-20 https://ineshin.space/papers/fused-linear-cross-entropy-apple-gpus/ru.html2026-07-19 https://ineshin.space/papers/fused-linear-cross-entropy-apple-gpus/nl.html2026-07-19 https://ineshin.space/papers/fused-linear-cross-entropy-apple-gpus/2026-07-19