I-Anthropic imemezele namuhla (mhla ziyi-8 kuNcwaba) ukuthi kusukela mhla ziyi-14 kuNcwaba izoshintsha imodi yezimvume ezenzakalelayo ku-Claude Code ibe yi-Auto Mode kubabhalisile be-Pro, Max ne-Team.
Ngokombhalo webhulogi ocashunwe yi-IT之家, i-Auto Mode isho ukuthi i-Claude Code isebenzisa i-classifier ezimele ye-AI ukuhlola ngesikhathi sangempela yonke i-tool call nomyalo we-Shell, ivumele imisebenzi ephephile futhi ivimbe ngokuzenzakalelayo izenzo ezilimazayo, ezingenakuhlehliswa noma ezeqa izimvume.
Mayelana nobubanzi bokusetshenziswa, i-Claude Enterprise, i-Claude API, i-Claude Platform on AWS, i-Amazon Bedrock, i-Google Cloud's Agent Platform ne-Microsoft Foundry zisazogcina ukukhetha imodi. I-Anthropic ihlela ukuqhubekisela phambili lolu shintsho kulezi zinkundla phakathi nenyanga ezayo, yenze i-Auto Mode ibe inketho ezenzakalelayo futhi iqede izindleko ezihambisana nayo.
Esivivinyweni, i-Anthropic yamema abahloli abakhokhelwayo abayi-1,053. Kutholakale ukuthi kwimodi evamile yokugunyazwa ngesandla kwabonakala kuphela u-13.6% wemiyalo eyingozi, kanti i-Auto Mode yakwazi ukubona u-89% wemiyalo eyingozi.
Ocwaningweni lokujova imiyalo, i-Trajectory Labs yasebenzisa izimo ezingu-72, yenza izivivinyo eziyi-10 ngesimo ngasinye, kwaba nemizamo yokuhlasela engu-720 isiyonke, ihlanganisa i-Claude Code v2.1.205 ne-Codex v0.144.5.
I-Claude Fable 5, i-Opus 5 ne-Sonnet 5 zonke zavimba ukuhlasela kwimodi ezenzakalelayo, ngaphandle komzamo owaphumelela; i-GPT-5.6 Sol yaba nezinga lempumelelo elingu-5.83% kwimodi ye-Auto-review ku-Codex, kwathi kwimodi ye-Full Access laba ngu-19.03%.
I-Anthropic iphinde yaveza ukuthi imodi ye-bypassPermissions ngaphandle kokuvikelwa okwengeziwe ibe nezinga elijwayelekile lempumelelo yokuhlasela elingu-0.09%, kanti inguqulo entsha ye-Auto-review ekhishwe yi-OpenAI ngesonto eledlule ingase iguqule imiphumela.
