[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-why-minimax-m27-self-evolution-matters-zh":3,"article-related-why-minimax-m27-self-evolution-matters-zh":35,"series-model-release-a553de67-f015-4f68-af27-06d57ab1e6b3":85},{"id":4,"title":5,"content":6,"summary":7,"source":8,"source_url":9,"author":10,"image_url":11,"keywords":12,"language":18,"translated_content":10,"views":19,"is_premium":20,"created_at":21,"updated_at":21,"cover_image":11,"published_at":22,"rewrite_status":23,"rewrite_error":10,"rewritten_from_id":24,"slug":25,"category":26,"related_article_id":27,"status":28,"google_indexed_at":10,"x_posted_at":10,"tweet_text":10,"title_rewritten_at":10,"title_original":10,"key_takeaways":29,"topic_cluster_id":33,"embedding":34,"is_canonical_seed":20},"a553de67-f015-4f68-af27-06d57ab1e6b3","為什麼 MiniMax M2.7 的自我進化，比基準分數更重要","\u003Cp data-speakable=\"summary\">\u003Ca href=\"\u002Fnews\u002Fminimax-m2-open-source-agentic-coding-zh\">Mini\u003C\u002Fa>Max M2.7 的重點不是分數，而是把模型改進變成代理式工作流。\u003C\u002Fp>\u003Cp>MiniMax M2.7 值得關注，不是因為它又多拿了幾個榜單分數，而是因為它把「模型如何變強」這件事本身，變成了代理式工作流。官方說法很清楚：模型參與自我演化、建立複雜 \u003Ca href=\"\u002Ftag\u002Fagent\">agent\u003C\u002Fa> harness、更新記憶，並在訓練與任務交付流程中反覆迭代。這代表競爭焦點正在改變，\u003Ca href=\"\u002Fnews\u002Fai-weekly-2026-w21-zh\">AI\u003C\u002Fa> 不再只是比誰的回答更像樣，而是比誰能把改進速度做成系統能力。\u003C\u002Fp>\u003Ch2>第一個論點\u003C\u002Fh2>\u003Cp>自我進化是對的戰略目標，因為真正稀缺的不是單次推理，而是持續改進的迴圈。MiniMax 描述的內部流程裡，M2.7 會更新 memory、建立技能、修改 harness，並在反覆實驗中調整結構。官方甚至舉例說，模型經歷超過 100 輪自動分析、scaffold 調整、評估與 rollback 決策後，內部評測提升了 30%。這不是小修小補，而是把模型進步從一次性訓練，改成工程化的迭代流程。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779077631921-lt4p.png\" alt=\"為什麼 MiniMax M2.7 的自我進化，比基準分數更重要\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>這件事重要，因為 AI 的瓶頸早就不只是算力或參數，而是人力協作成本。研究、基礎設施、評測、產品，四個團隊都會碰同一套系統，每次交接都拖慢迭代。MiniMax 宣稱 M2.7 在部分內部研究情境中可承擔 30% 到 50% 的工作流，若屬實，優勢就不是單純省時間，而是讓改進速度產生複利。能幫忙設計實驗、檢查失敗、提出修正的模型，價值遠高於只會回覆 prompt 的模型。\u003C\u002Fp>\u003Ch2>第二個論點\u003C\u002Fh2>\u003Cp>最能驗證代理式能力的地方，是軟體工程，而不是聊天。MiniMax 公布 M2.7 在 SWE-Pro 得到 56.22%，在 VIBE-Pro 得到 55.6%，在 Terminal Bench 2 得到 57.0%。這些都不是玩具任務，而是對 repo 級推理、除錯、系統理解與端到端交付的考驗。它還聲稱把部分線上事故的恢復時間壓到三分鐘內，做法包括觀測分析、資料庫檢查與根因推理。對工程團隊來說，這比單純的 code completion 更有意義，因為它碰到的是實際生產環境。\u003C\u002Fp>\u003Cp>更關鍵的是，MiniMax 把 M2.7 定位成系統模型，不是 code generator。它強調 agent team、角色邊界、對抗式推理、協議遵循與動態工具搜尋。這比「更會寫程式」的敘事成熟得多，因為生產工作真正需要的是協調、修正、測試與在混亂限制下做判斷。若 M2.7 真能在 harness 裡完成除錯、重寫、驗證與交接，那它的價值就不只在程式碼產出，而是在工程工作的操作層。\u003C\u002Fp>\u003Ch2>第二個論點\u003C\u002Fh2>\u003Cp>辦公軟體場景也不是噱頭，因為企業導入 AI 最先卡住的，往往就是文件工作。MiniMax 說 M2.7 在 GDPval-AA 拿到 1495 的 ELO，為\u003Ca href=\"\u002Ftag\u002F開源模型\">開源模型\u003C\u002Fa>中最高，且能改善 Excel、PowerPoint、Word 的複雜編輯。這很重要，因為企業不是只買給工程師用 AI；真正能擴散的模型，必須能改文件、保格式、處理多輪修改，還要理解商務語境。很多組織裡，AI 系統最先失敗的地方，就是這些看似平凡的文書流程。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779077631623-rjgt.png\" alt=\"為什麼 MiniMax M2.7 的自我進化，比基準分數更重要\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>所以，任務交付比單點推理更重要。MiniMax 提到 40 多個複雜技能、每個都超過 2000 tokens，技能遵循率達 97%。這顯示它在追求穩定性，而不是只追求聰明一次。對辦公場景來說，這才是正確優先順序：一次很驚豔但下一次失手的模型，無法進入真實流程；能長時間維持角色、遵守指令、處理\u003Ca href=\"\u002Ftag\u002F長上下文\">長上下文\u003C\u002Fa>的模型，才會被嵌進企業工作系統。\u003C\u002Fp>\u003Ch2>反方可能怎麼說\u003C\u002Fh2>\u003Cp>懷疑者會說，這仍然是一篇廠商自述，範圍太大，風險也太高。自我進化聽起來漂亮，但驗證困難；benchmark 可以針對題型優化，內部提升可能只是在受控環境裡成立；自動化迴圈在 demo 中看起來很強，到了真實生產就可能被雜訊、\u003Ca href=\"\u002Fnews\u002Fwhy-distributed-computing-is-the-default-zh\">例外\u003C\u002Fa>與權限問題打回原形。再加上文中大量對標頂級閉源模型，整體敘事很容易被看成一場聲量競賽。\u003C\u002Fp>\u003Cp>這些疑慮成立，但不代表訊號無效。重點不是每個數字今天都能被第三方完全重現，而是 MiniMax 指向的方向對不對。答案是對的。產業正在從 chatbot 走向 agent，從靜態模型走向能檢查、修改、評估並改善工作流的系統。就算部分數字偏樂觀，核心論點仍然站得住腳：真正的優勢會來自代理式迭代，而不是單次榜單成績。\u003C\u002Fp>\u003Ch2>你能做什麼\u003C\u002Fh2>\u003Cp>如果你是工程師，不要只看 M2.7 的分數，直接測它能不能進入你的工作流：事故分流、修補草案、執行檢查、保留上下文、乾淨交接。如果你是 PM 或創辦人，別把重點放在 demo 漂不漂亮，應該看任務完成率與迭代速度。把評測設計成真實工作：多步除錯、文件修訂、repo 級變更、跨工具協作。會贏的公司，不是把模型當更聰明的聊天機器人，而是把它當工作系統的一部分。","MiniMax M2.7 的重點不是分數，而是把模型改進變成代理式工作流；這代表 AI 競爭正從榜單轉向系統能力。","www.minimax.io","https:\u002F\u002Fwww.minimax.io\u002Fnews\u002Fminimax-m27-en",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779077631921-lt4p.png",[13,14,15,16,17],"MiniMax M2.7","自我進化","代理式工作流","軟體工程","企業 AI","zh",0,false,"2026-05-18T04:13:26.004663+00:00","2026-05-18T04:13:25.89+00:00","done","9dcc5192-b3fa-45d0-b6fa-2f497ac6b846","why-minimax-m27-self-evolution-matters-zh","model-release","2edf849f-e3e8-4462-9b87-ddb89d492cdf","published",[30,31,32],"MiniMax M2.7 的關鍵價值在於自我進化與代理式迭代，而不是單純的 benchmark 成績。","軟體工程與辦公工作是驗證這類模型是否真的能進入生產流程的核心場景。","工程師、PM、創辦人都應該用真實工作流來評估模型，而不是只看榜單與 demo。","0ccb5d2e-69f1-4354-a3e0-cb370221cd95","[-0.004103663,0.023463726,0.035545047,-0.08740138,-0.00880735,-0.025473742,0.011064271,0.020019228,-0.007385205,0.008488194,0.0074905064,-0.0064625638,-0.0029110413,0.0050797644,0.11665032,0.05114561,0.0017103016,-0.007855256,-0.010288599,0.008571277,0.0023218242,0.012513371,-0.014818624,0.00304548,-0.017662466,0.0049001127,0.011449706,0.0046593864,0.03806845,0.013957594,0.0038819301,-0.01120651,0.008899301,0.019512605,-0.0044617043,0.012793757,0.042165585,0.013443957,-0.0064077564,-0.0061287563,-0.0103294235,-0.03356262,-0.0022121733,-0.031181138,-0.009619503,-0.01729383,0.013515514,-0.036248215,-0.0017544199,-0.04048248,-0.02443761,0.024415798,-0.0046493225,-0.12986562,0.010160362,0.0050272313,0.005854499,0.009654307,0.0040023956,-0.012519138,-0.021851854,0.0019957202,-0.02883758,-0.0034159373,0.00954308,-0.01032199,0.036592823,0.0026052496,0.022965955,-0.0008608014,-0.0113909,-0.0040664813,0.0047298493,-0.014666137,0.005719824,-0.024341043,-0.009842188,0.02451684,0.0044694752,0.0067902207,0.008080365,-0.009546847,0.0053527383,0.0018799631,0.01572742,0.013539763,0.00087763794,0.0027106002,0.03874499,-0.009784448,0.006095628,-0.0062878863,0.0014469805,0.0030802458,-0.011847539,-0.0069498653,-0.02394356,0.0013978782,-0.0013113393,0.017881613,-0.017571164,-0.019836543,0.013129326,-0.0014537543,-0.00061923964,-0.003307969,-0.011256671,-0.009455396,0.01335259,-0.0041973554,-0.0062315892,-0.006467204,-0.007894323,-0.0051987856,-0.012234628,-0.12202283,0.011094103,0.01204652,-0.007744256,0.011966237,-0.031899795,0.019678045,0.0044717295,0.05365076,0.013726968,0.014933391,0.0021482012,-0.0012604491,-0.014788021,0.025585676,-0.025334794,0.02765904,0.009947631,-0.008041773,-0.018427107,0.02792363,-0.009037144,-0.013778118,-0.026603177,-0.016728312,0.0009317034,0.028998032,-0.01195981,0.007345476,-0.03145089,-0.0017644631,-0.052198622,0.015035896,-0.0020694223,-0.022047838,0.025134439,-0.02015813,-0.013914531,-0.0056162598,0.046367668,-0.026127722,0.0033390503,-0.010509464,0.004320904,0.004376652,0.017316103,-0.0009780027,-0.02254616,-0.00443676,-0.019852739,0.008805526,-0.0047044735,0.026460087,-0.00024740997,0.028808272,0.00019699874,-0.027476735,-0.0011501716,0.002725411,-0.020465171,-0.013857325,-0.020219827,-0.007355083,0.010704357,-0.012316268,0.01556158,0.008351658,0.0010827982,0.020114787,-0.01010351,0.011713803,0.015100085,0.0019532174,0.017450128,0.023831602,-0.040744048,0.0144565655,0.012082843,-0.03212325,0.012285369,-0.022338958,9.3800336e-05,0.0008771095,0.0073058666,0.021782858,0.0070367022,-0.01661949,0.015623034,-0.018295176,0.0171459,-0.014571017,0.01701957,-0.018615227,-0.00849922,-0.01753491,-0.007195329,-0.004452178,0.00102511,-0.024308309,-0.016989395,-0.008335082,0.026063817,0.004951655,0.015570711,-0.019138165,0.02721535,0.006530162,-0.0009167102,0.009347469,-0.006597093,-0.014455006,-0.016185476,-0.01849846,-0.007728845,0.040132966,-0.0028992097,0.033155974,0.0069521987,0.020532439,-0.0017500872,-0.014465008,0.012122989,0.02022051,0.008791986,0.019933607,-0.015921276,0.00040215073,0.006182822,0.009577193,0.010968565,-0.041300293,0.022653947,-0.011458363,-0.032008916,-0.010267725,-0.023544494,-0.003638758,0.010205856,-0.02868481,0.014723768,-0.026602019,-0.011119297,0.025226332,-0.016243633,0.004906136,-0.011113036,0.017557628,-0.011252922,0.012818457,0.004717422,0.0045073177,0.011954982,0.0067493105,-0.007898051,-0.0034973489,-0.0071823928,0.0020045312,0.007422756,0.00039514652,0.013693627,-0.0018745356,-0.030591408,0.01096173,0.0023977377,-0.015445851,0.009899317,0.0031828433,0.012747511,0.010725389,0.010379535,-0.009730146,-0.017723508,0.0009379686,-0.00402609,-0.0037919795,0.017633032,0.017256118,0.020078275,-0.0037625665,-0.0046335775,-0.017454758,0.0015406421,0.008015049,0.026117878,0.039322857,-0.00033008913,-0.0013607411,-0.0062380745,0.08872191,-0.012725927,-0.0091186995,0.0007786733,0.0012183355,-0.019290674,-0.0091771865,-0.023210041,-0.035881452,-0.006194181,-0.0037367027,-0.015963921,0.006942669,6.444822e-05,-0.033626556,0.003199612,-0.017488912,0.022200106,-0.021484617,-0.017048892,-0.0015866484,-0.004819537,-0.013857048,0.022245519,-0.007430799,-0.0073619285,-0.0072567905,0.0031457604,-0.0010018605,0.0378956,-0.01677389,0.001113889,-0.0006381625,0.018901052,0.005141962,-0.021897754,-0.0134340525,0.012382627,-0.0116034765,-0.011033814,-0.020468038,0.009958926,0.009291384,0.01211235,0.0067787496,0.012450672,-0.03327176,0.03912518,-0.013876139,0.010261254,-0.03549442,-0.008892452,0.035025604,0.019708836,-0.000667509,0.00767417,-0.01447225,0.0296882,-0.024899088,-0.0022488849,-0.0037459955,0.02752864,-0.018450327,0.005173836,0.027951814,-0.0028078873,-0.00025013223,0.008698449,0.007902489,0.0068573924,-0.0021138766,0.030718366,-0.0077212434,-0.017865598,-0.017620593,0.0068746363,0.0022019336,0.00021952824,0.037533622,-0.020938799,-0.013248189,-0.027784185,0.005248193,0.016588716,-0.015613878,0.0064176233,-0.009141742,-0.006136337,0.00819684,0.02135826,-0.009487781,0.019546423,0.0006628627,0.012808676,0.019799177,-0.024534818,0.0020702241,-0.0031536845,-0.0010314005,0.012034021,0.015481876,-0.022795752,0.0075347717,-0.01661454,0.011055266,-0.01455958,-0.01209995,0.00032726844,0.003069981,0.0011035909,0.0029536076,-0.015047946,-0.010676501,-0.013306788,0.00013515638,0.015809998,0.006679572,-0.014641333,-0.00037832028,0.025957016,0.026315555,0.008662122,0.014525947,0.010163829,0.027113805,0.024676414,-0.007133048,0.0103608435,-0.015980372,-0.019730967,-0.011827741,-0.009331107,-0.023322854,-0.020447167,-0.0012438738,-0.0068479874,-0.0020723639,-0.009591272,-0.007950876,0.01090609,-0.004446452,0.008373893,-0.0036626202,-0.020164821,-1.0293481e-05,-0.03137609,0.017738752,0.006246337,0.032199994,0.024923274,-0.009309544,-0.03645001,0.0026024978,-0.019706385,0.012442755,-0.012765415,-0.010297942,0.028380772,-0.01934578,-0.007874298,-0.013457364,0.003633967,-0.0072332784,9.158235e-05,-0.0051416303,-0.024918733,0.0026090313,0.0064357887,0.009027129,0.055066362,0.015021474,-0.008470865,-0.0071115764,-0.0014149387,0.002565385,-0.0024912197,-0.019724615,0.01857708,-0.01502889,0.014269936,-0.025193917,0.036549255,-0.0040404275,-0.04046486,-0.0036328887,-0.011446763,0.023419168,0.023543954,0.0071504815,0.0005427994,0.02811549,-0.011420262,0.00027900914,-0.00094498816,-0.035182394,0.01731506,0.004884266,-0.006464821,-0.011045232,0.013419007,-0.022455089,0.00429881,-0.012590392,0.007834036,-0.019339304,-0.018278284,-0.018424923,-0.02140511,0.004818338,0.012387485,0.030265657,-0.03466603,0.00051284774,0.025587967,-0.008763279,-0.015933951,-0.0003495938,0.0047737593,-0.02211686,-0.00444299,-0.015545092,-0.012003201,-0.006671886,0.018767964,0.025219418,0.013591108,0.03221374,-0.010199265,0.04743583,0.010700901,-0.00902494,-0.0011986282,-0.008164424,0.02910863,0.025264721,-0.017092783,0.0029437975,-0.0073927706,-0.0129499165,-0.009284516,-0.034125835,0.021212373,-0.10796088,-0.026748113,-0.004123662,-0.012796833,-0.01803269,0.004472786,0.0049514393,-0.011187736,0.010695236,-0.006351448,0.0034845907,0.0012705569,0.022526832,0.0015889377,0.004868021,-0.019859863,0.0042605596,0.0028691362,0.005788376,-0.020009136,0.0212798,0.027787358,0.017487466,-0.0019698895,0.012669782,-0.021712476,-0.0030394115,-0.011143941,0.020739755,-0.021699604,-0.01773732,-0.008840697,0.021191252,-0.0082860645,0.0343916,-0.0073252143,0.020180637,-0.0077304756,0.0015701583,0.0001011027,0.006921747,0.020468755,-0.030507073,-0.035499122,0.010293632,-0.0016053369,-0.03895756,0.008204643,0.00025254893,0.0004456622,-0.03296529,-0.015050412,-0.01421384,-0.01968409,-0.023923937,-0.0008536069,-0.025916822,-0.004122717,0.00595678,0.008577886,-0.014832101,-0.022947032,-0.0117842425,0.0051480597,-0.027886877,0.0051935404,-0.009180419,0.008769916,-0.002285294,0.023908973,-0.026224572,-0.022149349,0.007045046,0.0040014186,-0.009308234,0.004531853,-0.02221898,0.017676773,-0.013778803,-0.002771571,-0.03560397,-0.013955676,-0.10829032,-0.008311353,0.0069446703,-0.018268153,0.011504132,-0.0122151645,0.0029297546,-0.019843364,-0.020922394,0.004108854,0.026404157,0.01468834,-0.021305578,-0.03580426,-0.017840756,0.0018780068,0.020015394,0.019631306,0.0025822222,-0.019224335,-0.01696372,0.008669195,0.015276098,-0.026007483,-0.002560258,0.014121758,-0.018154785,0.01379237,-0.0029488825,0.025466396,-0.02197414,-0.1278522,-0.010590728,0.0038523653,-0.01975942,0.0007284811,-0.008876331,0.014587075,-0.0052952752,-0.0117038265,0.009087866,0.015048499,-0.010800963,-0.013622315,0.012171656,0.0016222637,0.14207447,0.0030546247,-0.02216032,-0.027749695,-0.0116205355,0.007395044,-0.0050161635,0.004056263,0.006751269,0.0100571625,-0.010449554,0.026162684,0.009577451,0.013404581,0.0059629725,0.0089459885,0.011240657,-0.004550827,-0.014089786,0.015705578,0.013527926,-0.01634849,-0.048264414,0.009517087,0.0023428989,0.0076214694,0.027597371,0.0176603,-0.0053877295,-0.009823778,0.01226846,0.007935417,0.008611517,-0.005613545,0.008960284,-0.008903199,-0.09230913,-0.0034394949,-0.02841072,-0.0011899023,0.0029540092,-0.014721748,-0.0019793452,0.0048024566,0.012452169,0.001992084,0.0077209016,0.013351319,0.04494651,-0.0106958095,0.020948414,0.036253083,-0.010127514,0.005720746,0.018095383,0.009004113,-0.0005610908,-0.018447097,-0.008022531,-0.012829937,-0.043130953,0.014596882,0.02346179,0.015325009,-0.020347431,0.013108179,-0.007249045,0.01449026,0.0021491896,-0.009317949,0.019560795,0.027599601,0.017292602,-0.0036930644,-0.0009745874,-0.011328378,0.00825872,-0.008553421,0.031460457,0.019722173,0.038272306,0.007213545,-0.0068529793,-0.014015021,-0.014838997,0.008492968,-0.0021460494,0.025789449,-0.004606701,-0.018672144,0.028151777,-0.011783969,0.042089716,0.022748452,0.004712586]",{"tags":36,"relatedLang":44,"relatedPosts":48},[37,38,40,42,43],{"name":15,"slug":15},{"name":17,"slug":39},"企業-ai",{"name":13,"slug":41},"minimax-m27",{"name":16,"slug":16},{"name":14,"slug":14},{"id":27,"slug":45,"title":46,"language":47},"why-minimax-m27-self-evolution-matters-en","Why MiniMax M2.7’s Self-Evolution Claim Matters More Than Its Benchma…","en",[49,55,61,67,73,79],{"id":50,"slug":51,"title":52,"cover_image":53,"image_url":53,"created_at":54,"category":26},"f861f009-eb92-4b55-bf2c-a4671c660f8a","minimax-m2-open-source-agentic-coding-zh","MiniMax M2 開源，代理編碼變便宜","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779077043926-exf5.png","2026-05-18T04:03:36.718257+00:00",{"id":56,"slug":57,"title":58,"cover_image":59,"image_url":59,"created_at":60,"category":26},"376052ab-c6d7-4efd-a4fc-d3c0b64c8997","copilot-studio-gpt-41-default-gpt-4o-retirement-zh","Copilot Studio 預設改用 GPT-4.1","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779031444364-txpm.png","2026-05-17T15:23:33.056197+00:00",{"id":62,"slug":63,"title":64,"cover_image":65,"image_url":65,"created_at":66,"category":26},"eee9cee3-8c3d-48a7-b647-98cf19955e54","claude-api-model-guide-new-top-tier-zh","Claude API 模型指南升級","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779023645425-mkdf.png","2026-05-17T13:13:36.325561+00:00",{"id":68,"slug":69,"title":70,"cover_image":71,"image_url":71,"created_at":72,"category":26},"c2abd58c-029c-4e1e-97cc-8f5a5ca969e2","mistral-cybersecurity-model-banks-europe-zh","Mistral 要做銀行資安模型","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778999023148-7ikg.png","2026-05-17T06:23:22.680619+00:00",{"id":74,"slug":75,"title":76,"cover_image":77,"image_url":77,"created_at":78,"category":26},"cf56a4be-6a68-4053-b262-0b343406b995","kimi-k2-6-complete-guide-2026-zh","Kimi K2.6 2026 變了什麼","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778981033614-xldp.png","2026-05-17T01:23:31.568906+00:00",{"id":80,"slug":81,"title":82,"cover_image":83,"image_url":83,"created_at":84,"category":26},"da646ee1-db2e-48b4-9ffe-b79a8a81ae93","why-kimi-k26-changes-coding-model-race-zh","為什麼 Kimi K2.6 會改寫寫程式模型競賽","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778980423319-9xts.png","2026-05-17T01:13:21.45526+00:00",[86,91,96,101,106,111,116,121,126,131],{"id":87,"slug":88,"title":89,"created_at":90},"58b64033-7eb6-49b9-9aab-01cf8ae1b2f2","nvidia-rubin-six-chips-one-ai-supercomputer-zh","NVIDIA Rubin 把六顆晶片塞進 AI 機櫃","2026-03-26T07:18:45.861277+00:00",{"id":92,"slug":93,"title":94,"created_at":95},"0dcc2c61-c2a6-480d-adb8-dd225fc68914","march-2026-ai-model-news-what-mattered-zh","2026 年 3 月 AI 模型新聞重點","2026-03-26T07:32:08.386348+00:00",{"id":97,"slug":98,"title":99,"created_at":100},"214ab08b-5ce5-4b5c-8b72-47619d8675dd","why-small-models-are-winning-on-device-ai-zh","小模型為何吃下裝置端 AI","2026-03-26T07:36:30.488966+00:00",{"id":102,"slug":103,"title":104,"created_at":105},"785624b2-0355-4b82-adc3-de5e45eecd88","midjourney-v8-faster-images-higher-costs-zh","Midjourney V8 變快了，也變貴了","2026-03-26T07:52:03.562971+00:00",{"id":107,"slug":108,"title":109,"created_at":110},"cda76b92-d209-4134-86c1-a60f5bc7b128","xiaomi-mimo-trio-agents-robots-voice-zh","小米 MiMo 三模型瞄準代理、機器人與語音","2026-03-28T03:05:08.779489+00:00",{"id":112,"slug":113,"title":114,"created_at":115},"9e1044b4-946d-47fe-9e2a-c2ee032e1164","xiaomi-mimo-v2-pro-1t-moe-agents-zh","小米 MiMo-V2-Pro 登場：1T MoE 模型","2026-03-28T03:06:19.002353+00:00",{"id":117,"slug":118,"title":119,"created_at":120},"d68e59a2-55eb-4a8f-95d6-edc8fcbff581","cursor-composer-2-started-from-kimi-zh","Cursor Composer 2 其實從 Kimi 起步","2026-03-28T03:11:58.893796+00:00",{"id":122,"slug":123,"title":124,"created_at":125},"c4b6186f-bd84-4598-997e-c6e31d543c0d","cursor-composer-2-agentic-coding-model-zh","Cursor Composer 2 走向代理式寫碼","2026-03-28T03:13:06.422716+00:00",{"id":127,"slug":128,"title":129,"created_at":130},"45812c46-99fc-4b1f-aae1-56f64f5c9024","openai-shuts-down-sora-video-app-api-zh","OpenAI 關閉 Sora App 與 API","2026-03-29T04:47:48.974108+00:00",{"id":132,"slug":133,"title":134,"created_at":135},"e112e76f-ec3b-408f-810e-e93ae21a888a","apple-siri-gemini-distilled-models-zh","Apple Siri 牽手 Gemini 的真相","2026-03-29T04:52:57.886544+00:00"]