[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"tag-記憶體瓶頸":3},{"tag":4,"articles":9},{"id":5,"name":6,"slug":6,"article_count":7,"description_zh":8,"description_en":8},"e2450263-7d98-4bb8-8f46-b4c3cf9969d4","記憶體瓶頸",0,null,[10,19],{"id":11,"slug":12,"title":13,"summary":14,"category":15,"image_url":16,"cover_image":16,"language":17,"created_at":18},"ad2e19d7-a96f-4a39-bd32-5b139f46b560","why-turboquant-matters-more-than-model-size-zh","為什麼 TurboQuant 比模型大小更重要","TurboQuant 之所以重要，不是因為模型更大，而是因為它直接壓低了決定本地 AI 表現的 KV cache 記憶體瓶頸。","research","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779864485542-p489.png","zh","2026-05-27T06:47:24.622955+00:00",{"id":20,"slug":21,"title":22,"summary":23,"category":24,"image_url":25,"cover_image":25,"language":17,"created_at":26},"3c206419-ad56-478e-a9d4-203832c11744","why-kv-cache-compression-will-decide-edge-ai-inference-zh","為什麼 KV-cache 壓縮會決定邊緣 AI 推論","我認為邊緣 AI 推論的勝負，不會先由算力決定，而是由 KV-cache 壓縮這個記憶體瓶頸決定。","tools","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779285832259-zgfd.png","2026-05-20T14:03:19.991728+00:00"]