Izakhono ze-Anthropic's Agent: umgangatho omtsha ovulekileyo wee-arhente ze-AI kwishishini

Ishishini lobukrelekrele bokwenziwa kweshishini lijongene nenyikima encinci ngenxa yentshukumo ye I-Anthropic kunye nesiphakamiso sayo sezakhono ze-AgentKunokuba ikhuphe olunye uphawu oluvaliweyo, inkampani ikhethe ukupapasha iinkcukacha ezivulekileyo ezithi Ivumela nayiphi na intlangano ukuba ichaze, yabelane, kwaye ilawule izakhono ze-AI ngendlela emiselweyo.Oku kubaluleke kakhulu kwiinkampani zaseYurophu ezisebenza kwiindawo ezilawulwayo.

Enyanisweni, oku kuthetha ukuba abancedisi be-AI bayayeka ukuthembela kwizicelo ezizenzekelayo baze baqale ukusebenza nabo iilayibrari zezakhono ezicwangcisiweyo, eziguqulelweyo, nezinokuphicothwaenokusetyenziswa kwakhona kumaqela amaninzi, usetyenziso, kunye nabathengisi. Kwiinkampani eSpain nakwamanye amazwe aseYurophu sele zivavanye iiarhente ze-AI kwezomthetho, kwezemali, okanye kwinkonzo yabathengi, le ndlela isetyenziswa Ithembisa ulawulo olungakumbi, "umlingo omnyama" omncinci, kunye nokuhlanganiswa okucwangcisiweyo neenkqubo zayo zangaphakathi..

Yintoni i-Agent Skills kwaye kutheni iphawula utshintsho olukhulu kwi-AI yeshishini?

Izakhono ze-Agent, ngokuyintloko, zizinto isakhelo esiqhelekileyo sokufundisa iiarhente ze-AI imisebenzi ethile yomsebenziUlwazi lufakwe kwiimodyuli ezizimeleyo. Izakhono zonke ziyifolda okanye iphakheji enemiyalelo yenyathelo ngenyathelo, izikripthi, imizekelo yokusetyenziswa, kunye nezixhobo ezithile ezixelela iimodeli ezifana noClaude indlela yokwenza kwimeko ethile yobungcali: ukuvelisa ingxelo yezemali ngokwemigaqo, ukulungiselela intetho enezikhokelo zophawu, okanye ukucubungula imbuyekezo ngokwemigaqo-nkqubo yenkampani.

Endaweni yendlela eqhelekileyo "yokubuza izinto" kwimodeli enezindululo ezinde, imibutho inokudala iingqokelela zangaphakathi zezakhono ezibonisa iinkqubo zazo zokwenyaniEzi layibrari zabelwana ngazo zonke iiqela, zijongwa ngokungathi ziyikhowudi, kwaye zifakwe kwizixhobo ezisetyenziswa imihla ngemihla. Kwiinkampani ezininzi zaseYurophu, le ndlela ihambelana ngcono neemfuno zazo zokuthobela imithetho, ulawulo lwedatha, kunye nokulandelelwa kwayo.

Olunye utshintsho olubalulekileyo kukuba i-Anthropic ayipheleli ekusebenziseni i-Agent Skills ngaphakathi kwenkqubo yayo yendalo: Inkcazo ipapashwe njengomgangatho ovulekileyo.Oku kufana noko kwenziwe yinkampani ngeModel Context Protocol yayo (MCP), ngoku eyamkelwe ngokubanzi ekudibaniseni iiarhente neenkonzo zangaphandle. Nawuphi na umboneleli, nokuba yinkampani enkulu yelifu okanye inkampani yesoftware ethile kushishino lwe-EU, unokuphumeza nokwandisa umgangatho ngaphandle kokubotshelelwa kumthengisi omnye.

Kwimarike apho iimodeli ezivela kwi-OpenAI, Google, Anthropic, kunye nabanye abadlali zihleli kunye, zine ulwimi oluqhelekileyo ukuchaza izakhono zeearhente Ijolise ekunciphiseni ukuxhomekeka kumaqonga abucala kunye nokwenza lula ukufuduka okanye ukusasazwa kwe-hybrid, into exatyiswa ngakumbi ziibhanki zaseYurophu, abanini bomshuwalense okanye ulawulo lukarhulumente.

Inqaku elinxulumene nalo:

I-Nemotron 3: Ubhejo olukhulu lwe-NVIDIA oluvulekileyo kwi-AI yeearhente ezininzi

Indlela i-Agent Skills esebenza ngayo kwaye zeziphi iingxaki ezizisombululayo

Izakhono ze-Agent ziboniswa njenge iimodyuli ezihlanganisiweyo ezihlala phakathi komzekelo wolwimi kunye neenkqubo zangaphakathiLe modeli iseyiyo eqondayo, eqiqayo nethetha izinto eziphikisanayo, kodwa xa kufuneka "yenze" izinto ezibonakalayo—jonga ibhalansi, uvule ithikithi eJira, wenze ingxelo yomthetho— isebenzisa isakhono esifanelekileyo, esichaza ngokuchanekileyo indlela yokuqhubeka.

Isakhono ngasinye sidla ngokuquka ifayile yenkcazo (njengesaziwayo SKILL.mdEli candelo lichaza, ngendlela exutyiweyo ye-YAML kunye nombhalo ocwangcisiweyo, igama lesakhono, amanyathelo ekufuneka alandelwe, iiparameter ezivumelekileyo, imizekelo yokusetyenziswa, kunye nezixhobo okanye ii-API ezinokubizwa. Akukho manyathelo afanelekileyo ashiyelwa kwingozi: Zisetyenziswa njengekhowudi emiselweyo ebiza iinkonzo zoshishinongelixa imodeli igxile kwiinkalo zencoko kunye nokwenza izigqibo.

Ukuphucula ukusebenza kakuhle, i-Anthropic ifake uyilo lwe "ukutyhilwa okuqhubekayo"Umncedisi akalayishi zonke iinkcukacha zesakhono ngasinye ngokwemeko; ufikelela kuphela kulwazi olupheleleyo xa lufuneka ngokwenene. Ngale ndlela, umbutho unokugcina ithala leencwadi elikhulu kakhulu ngaphandle kokugqithisa imemori yomzekelo, nto leyo eluncedo ngakumbi kwiindawo ezinzima ezifana neebhanki, ii-telecoms, okanye abathengisi abakhulu baseYurophu.

Elinye icandelo eliqhelekileyo yileyo ibizwa ngokuba iarhente yokucwangcisa, loo nto usebenza njengomphathi: ufumana isicelo somsebenzisi, ufumanisa injongo, ugqiba kwelokuba loluphi udibaniso lwezakhono nezixhobo ezifunekayo aze azilandelelaniseUmbuzo olula wokuhlawula unokubangela isakhono sokucacisa injongo, isakhono "sokuchaza i-invoyisi yam", kwaye, ngaphantsi, isixhobo esibuza iinkqubo zokuhlawula ngaphandle kokuba umsebenzisi aqonde ubunzima.

Kule ndlela, izakhono ziba indlela yokuphumeza iiarhenteInqanaba lencoko lihlala liguquguquka, ngelixa iinkqubo zichazwa, zingasetyenziswa kwakhona, kwaye zilawulwa ngumgangatho. Ilungisa enye yeziphene eziphambili zeerobhothi kunye nabancedisi bokuqala abasebenzisa i-AI, ababenzima ukuyihlola indlela abaziphethe ngayo. kwaye yatshintsha ngendlela engalindelekanga xa izindululo ziguqulwa.

Ukuvuleleka, umgangatho, kunye nokwamkelwa kwangoko kwenkqubo yendalo

Eyona nto imangalisayo eyenziwa yi-Anthropic kukupapasha Iinkcukacha zobugcisa ze-Agent Skills kunye ne-SDK yayo njengomgangatho ovulekileyo ngokusebenzisa i-agentskills.io, bemema uluntu kunye nabanye ababoneleli ukuba bayamkele baze bayiphuhlise. Eli nyathelo lilandela i-MCP, esandula ukuba phantsi kolawulo lwe Isiseko seLinux ngaphakathi kwe Agetic AI Foundation, apho abadlali abafana ne-AWS, iGoogle, iMicrosoft okanye iBlock bathatha inxaxheba khona.

Malunga nezakhono ze-Agent, ukwamkelwa kwangoko ziinkampani ezinkulu zobuchwephesheIzixhobo ezifana neMicrosoft VS Code, iGitHub, kunye neearhente zokubhala iikhowudi ezifana neCursor kunye ne-OpenCode zifake izakhono zokwakha ukuchaza iindlela zophuhliso. I-OpenAI ngokwayo ingenise izakhiwo ezifanayo kakhulu kwiChatGPT kunye ne-CLI yomphuhlisi wayo, kunye nee-directory zezakhono ezikhumbuza indlela ka-Anthropic, nto leyo ebonisa ukudibana okuthile ngaphakathi kwishishini ukuya kolu hlobo lwe-modularity.

Okwangoku, iinkampani eziphambili zesoftware yeshishini —I-Atlassian, i-Figma, i-Stripe, i-Canva, i-Notion, i-Cloudflare, i-Zapier okanye i-RampIinkampani ezifana ne-[igama lenkampani] zipapasha izakhono zazo zokudibanisa iimveliso zazo neearhente ze-AI. Ezi zakhono zivumela abasebenzisi ukuba, umzekelo, benze imisebenzi kwiJira okanye kwiTrello emva kwemigaqo yangaphakathi, basebenzise izitayile zebhrendi kuyilo lweFigma, okanye benze ngokuzenzekelayo imisebenzi yokuthengisa ngaphandle kokufuna ukuhlanganiswa okungacwangciswanga kumthengi ngamnye.

Uluntu lwabaphuhlisi nalo luyabandakanyeka: Indawo yokugcina izakhono ze-Anthropic iqokelele amashumi amawaka eenkwenkwezi kwiGitHub nakwi Sele kukho amawakawaka ezakhono ezabelwana ngazo esidlangalaleni, ukusuka kwizinto eziluncedo zokuguqula iiPDF ukuya kwizinto ezithile ezizenzekelayo zobunjineli okanye amaqela ezemali.

Le nkqubo yendalo inomdla ngakumbi kwiinkampani zaseYurophu ezisebenzisa kakhulu izixhobo ezifana ne-Atlassian, iMicrosoft 365 okanye i-Figma kwaye zifuna iiarhente zazo ze-AI zisebenze nazo ngelixa zihlonipha imigaqo-nkqubo yangaphakathi, imigaqo yecandelo kunye neemfuno zabucala ezifana ne-GDPR. ngaphandle kokuxhomekeka kwizandiso ezingacacanga ezivela kumboneleli omnye.

Ukusuka kwisixhobo somphuhlisi ukuya kwiziseko zophuhliso lweshishini

Xa i-Anthropic yazisa ezi zakhono ngo-Okthobha, izakhono zazibonwa ikakhulu njenge Isixhobo esiluncedo kubaphuhlisi kunye nabathandi bekhowudiNgokusebenzisa "umdali wezakhono" osebenzisanayo eClaude, abasebenzisi ngokwabo banokuvelisa ulwakhiwo lwefolda kunye ne-SKILL.md efunekayo ukuze kusebenze ngokuzenzekelayo imisebenzi ethile, ngaphandle kokusetyenziswa kobunjineli obukhulu.

Ngohlaziyo lwakutshanje, inkampani itshintshele ingqalelo yayo kwishishini: Izakhono ze-Agent ngoku zidibene izixhobo zolawulo lombuthoIsikhokelo esiphambili sezakhono kunye nemisebenzi yolawulo eyenzelwe abaphathi be-IT kunye namaqela okhuseleko. Ingcinga kukuba izakhono zidlulele ngaphaya kovavanyo olusasazekileyo kwaye zibe ziimpahla ezizinzileyo, ezibhalwe phantsi, nezilawulwayo njengenxalenye yesiseko se-AI yeshishini.

Kwimibutho ebhalisele izicwangciso zeQela kunye nezeShishini likaClaude, izakhono zinokulawulwa ukususela iphaneli ephakathiKulapho abalawuli bagqiba khona ukuba zeziphi izakhono ezibonelelwa kwiqela ngalinye labasebenzisi, ezivunyiweyo ngokungagqibekanga, kwaye zeziphi ezifuna ukungena. Olu luhlu lolawulo luvumela ukulungelelanisa ukusetyenziswa kweearhente nemigaqo-nkqubo yangaphakathi, nto leyo ibalulekileyo kumacandelo alawulwa kakhulu eYurophu, afana nokhathalelo lwempilo, i-inshurensi, kunye neebhanki.

Ukongeza, i-Anthropic ivule i- Uluhlu lwezakhono zamaqabane oshishino Isebenza njengekhathalogu yezakhono ezilungele ukusetyenziswa, kunye negalelo elivela kwiinkampani ezifana ne-Atlassian, i-Canva, i-Figma, i-Notion, i-Cloudflare, i-Stripe, i-Zapier, kunye ne-Sentry. Kwiinkampani ezininzi zaseYurophu ezincinci kunye neenkampani ezinkulu, olu hlobo lwendawo yokugcina izinto luyenza lula iiprojekthi zovavanyo: endaweni yokwakha yonke into ukusuka ekuqaleni, bangaqala ngezakhono ezivavanyiweyo kwangaphambili baze bazilungelelanise neenkqubo zabo.

Konke oku kuthetha ukuba, ngaphezu nje kwento eveliswayo, i-Agent Skills iyaguquka ibe yi- umaleko weziseko zophuhliso apho uzakwakha khona iiarhente ze-AI kunye nezicelo, ngokuhambelana noko kwakuthethwa kukubekwa emgangathweni kwee-API ngelo xesha: ulwimi oluqhelekileyo apho izixhobo ezahlukeneyo zinokusebenzisana khona.

Imveliso, iimeko zokusetyenziswa kunye neenzuzo zeenkampani zaseYurophu

Ukuqaliswa kokuqala kwehlabathi lokwenyani kubonisa ukuba ukwamkelwa kwezakhono ze-Agent akusiyonto nje eqikelelwayo. Amaqela obunjineli axele ukuba imveliso inyuke ukuya kuthi ga kwi-50%. ngenxa yokwenza imisebenzi ephindaphindwayo ngokuzenzekelayo kunye nokulungiswa kwemisebenzi efana nokuhlolwa kwekhowudi, amaxwebhu obuchwephesha, okanye ukuveliswa kovavanyo.

Kwicandelo lezemali neleze-akhawuntingi, izakhono zivumela bhala phantsi iinkqubo ezilawulwayoUkususela kwiitshekhi ngaphambi kokukhupha ingxelo, ukuya kulawulo lokuthobela imithetho olusebenza ngokuzenzekelayo ngaphambi kokuba luvume iintengiselwano ezithile. Kwiinkampani zaseSpain eziphantsi kwemithetho yaseYurophu—njengeMiFID II yeenkonzo zotyalo-mali okanye iSolvency II yeinshorensi—ukukwazi ukuguqula le mithetho ibe zizakhono eziphicothwayo kuyinzuzo ngaphezu kwezicelo ezingacwangciswanga.

Kwimisebenzi nakwi-ofisi engasemva, imibutho isebenzisa amathala eencwadi ezakhono ukuze ukwabelana ngolwazi lwezikoOko kwakusaziwa ngaphambili ngabasebenzi abambalwa abanamava ngoku kuqulethwe kwiimodyuli ezinokuthi iarhente okanye umsebenzi omtsha azilandele inyathelo ngenyathelo, ukunciphisa ukuxhomekeka kubantu abathile kunye nokukhawulezisa uqeqesho lwangaphakathi.

Kuye kwavavanywa novavanyo oluneenjongo ezinkulu ngakumbi, olufana neprojekthi yangaphakathi ka-Anthropic yokulawula ivenkile encinci yokuthengisa impahla enabameli abaxhotyiswe ngezakhono kwimpahla, ukuthengisa, kunye nenkonzo yabathengi. Nangona ukujongwa kwabantu kusekho kwiimeko ezithile ezigqithisileyo, uvavanyo lubonisa ukuba Iiarhente ezixhotyiswe ngezakhono eziyilwe kakuhle zinokwenza imisebenzi yokuphela kwento kwiindawo ezilawulwayo.

Kwimeko yaseYurophu, apho iKhomishini kunye nabalawuli belizwe baqala ukufuna ukucaca okukhulu kunye nolawulo lweenkqubo ze-AILe ndlela yokulinganisa inceda uvavanyo lomngcipheko: isakhono ngasinye singabhalwa phantsi, sivavanywe, kwaye siqinisekiswe ngokuzimeleyo, ngelixa imodeli iyonke isetyenziswa njengengqiqo kunye nomgangatho wolwimi lwendalo.

Iingozi, ulawulo kunye nokungaqiniseki okujikeleze umgangatho

Ukuvula izakhono ze-Agent akunazo iingozi. Ngokuvumela nabani na ukuba athumele kwaye abelane ngezakhono, Kukho ithuba lokuba izakhono ezinobungozi okanye ezikumgangatho ophantsi zinokuvelangemiyalelo enokubangela iimpazamo, ukungathotyelwa kwemithetho, okanye ukuvuza kolwazi ukuba luqhagamshelwe kwiinkqubo ezinobuzaza.

I-Anthropic icebisa iinkampani ukuba Ukunciphisa ukwamkelwa kwezakhono kwimithombo ephicothiweyo kunye nabaphuhlisi abaqinisekisiweyokwaye badibanise uphononongo lwala mandla kwiinkqubo zabo zokhuseleko kunye nokuthobela imithetho rhoqo. Le nkampani ikwathatha inxaxheba kwiingxoxo noluntu malunga nokuba ngubani omele alawule ukuvela kwexesha elide kweprotocol evulekileyo kunye nendlela, umba obalulekileyo ukuba umgangatho uza kuthintelwa ukubanjwa ngumdlali omnye.

Enye ingxoxo eqhubekayo yimpembelelo izakhono zabantu kwimibuthoNjengoko iiarhente zizenzela ngokuzenzekelayo zonke iinkqubo, ezinye iingcali zilumkisa ngomngcipheko wezakhono "zokungakhathali": ukuba iqela liqhela i-AI ehlala ilungiselela iingxelo, ifaka amabango, okanye ilawula iinkqubo zenkonzo yabathengi, linokulahlekelwa bubuchule bokukwenza ngesandla xa kukho into engahambi kakuhle.

Abahlalutyi beli shishini bakwathi, nangona i-MCP iye yaba ngumgangatho oqhelekileyo, Akuqinisekwanga ukuba i-Agent Skills iya kuphinda impumelelo efanayoImibutho sele iqhele ukusebenza ngee-API ezisemgangathweni kunye neempawu zonxibelelwano, kwaye kukho iindlela ezininzi zokufundisa izakhono kwiiarhente. Ngamanye amazwi, iingenelo zobugcisa zeZakhono zeAgent zodwa azanele ukuqinisekisa ukwamkelwa ngokubanzi.

Kwiinkampani zaseYurophu, eziqhele ukusebenza kwiindawo ezithengisa abathengisi abaninzi, oku kuthandabuza kuthetha ukuba lulumko: uninzi luzama izakhono ze-Agent kwiiprojekthi zovavanyo, kodwa zigcina ngaxeshanye. amaqhinga athile okulungiselela nokulawula iiarhente, kunye neengqimba zolawulo ezingaphezulu kwayo nayiphi na imigangatho ethile.

Iingenelo zobuchule zabasunguli kunye nee-CTO zezinkampani ezintsha eSpain naseYurophu

Ngaphandle kweenkampani ezinkulu, i-Agent Skills ivula ithuba elinomdla kwi Iinkampani ezintsha zobuchwepheshe zaseYurophu kunye nokwandiswa kwazoKwamaqela amaninzi asekayo, umahluko wokwenyani awusasebenzisi nje "imodeli engcono kakhulu" kwimarike, kodwa usebenzisa ubuchule bawo ngendlela yezakhono zobunini ezibamba iinkqubo zabo, indlela abasebenza ngayo, kunye nokuqonda kwabo umthengi.

Ngale ndlela, ukutyala imali kwizixhobo zokwakha iilayibrari zezakhono ezimele ubukrelekrele bombutho Oku kunokuba yimpahla yexesha elide, efana nokuba ne-API eyilwe kakuhle okanye isiseko sedatha esomeleleyo. Ezi zakhono zinokusetyenziswa kwiimodeli ezahlukeneyo kunye namaqonga, ukunciphisa ukuxhomekeka kumthengisi othile kunye nokwenza kube lula ukuthotyelwa kweemfuno zaseYurophu malunga nolawulo lwedatha okanye indawo.

Umgangatho ovulekileyo ukwanceda ukusebenzisana phakathi kwezisombululo ezivela kubaboneleli abahlukeneyoIshishini elitsha laseSpain eliphuhlisa imveliso ye-SaaS, umzekelo, yolawulo lwamaxwebhu kwiifemu zomthetho, linokubonisa izakhono zalo njengezakhono ezihambelana noClaude, kodwa likwabonisa nezinye iiarhente ezisebenzisa iinkcukacha ezifanayo, ngaloo ndlela zandisa imakethi yalo ngaphandle kokuphinda lidibanise iqonga ngalinye.

Ngaphezu koko, inkqubo yeqabane—enezixhobo ezifana ne-Atlassian, i-Figma, i-Stripe, kunye ne-Zapier—inika iinkampani ezintsha indlela emfutshane: endaweni yokwakha izihlanganisi ezintsonkothileyo kwinkonzo nganye, zinokusebenzisa izakhono ezikhoyo kwaye zigxile kuzo. yongeza iileya zengqiqo kunye namava obuqu phezuluOku kuhambelana kakuhle nenyani yeenkampani ezininzi zaseYurophu, ezisebenza namaqela amancinci kwaye zifuna ukwandisa imbuyekezo kwisantya ngasinye sophuhliso.

Kwi-CTOs eziqala ukuyila isicwangciso seearhente zazo, isifundo sicacile: phatha izakhono njengezinto zexesha elideukuguqulela, ukubeka esweni, kunye nokuziphucula ngedatha yokwenyani, kunye nokuzilungelelanisa nomgangatho wolawulo nolawulo oluchazwa yintlangano. Ngale ndlela, xa inkqubo yendalo ikhula—kwaye imigangatho izinza—inkampani iya kuba sele inekhathalogu yayo yezakhono, ezilungele ukudityaniswa naphi na apho kufanelekileyo khona.

Ukuvulwa kwe-Anthropic kwe-Agent Skills kuphinda kuchaze indlela ii-arhente ze-AI ezithathwa ngayo kwishishini: ukusuka kubancedisi ngokubanzi abalawulwa ziingcebiso, ukuya amaqonga omsebenzi asekelwe kwizakhono eziziimodyuli, eziphathwayo, nezinokuphicothwaESpain naseYurophu, apho uxinzelelo lolawulo kunye nesidingo sokusebenzisana ziphezulu kakhulu, le modeli inika indlela ephakathi phakathi kobuchule obukhawulezayo kunye nolawulo oluqinileyo, ishiya ucango luvulekile lokuba ixabiso lokwenyani lokwahlula libe kwizakhono eziza kwakhiwa kwaye zilawulwe yintlangano nganye.

UAlberto Navarro

Ndingumntu othanda itekhnoloji ojike umdla wakhe we "geek" waba ngumsebenzi. Ndichithe ngaphezulu kweminyaka eli-10 yobomi bam ndisebenzisa itekhnoloji yokusika kwaye ndikhenkceza ngazo zonke iintlobo zeenkqubo ngenxa yokufuna ukwazi okumsulwa. Ngoku ndiqeqeshelwe ubugcisa bekhompyutha nakwimidlalo yevidiyo. Oku kungenxa yokuba ngaphezu kweminyaka emi-5 ndibhalela iiwebhusayithi ezahlukeneyo kwitekhnoloji kunye nemidlalo yevidiyo, ndisenza amanqaku afuna ukukunika ulwazi oludingayo ngolwimi oluqondakalayo kuye wonke umntu.

Ukuba unayo nayiphi na imibuzo, ulwazi lwam lusuka kuyo yonke into enxulumene nenkqubo yokusebenza yeWindows kunye ne-Android yeefowuni eziphathwayo. Kwaye ukuzinikela kwam kukuwe, ndihlala ndikulungele ukuchitha imizuzu embalwa kwaye ndikuncede usombulule nayiphi na imibuzo onokuba nayo kweli lizwe le-intanethi.