{"id":15695,"date":"2026-06-05T18:08:35","date_gmt":"2026-06-05T18:08:35","guid":{"rendered":"https:\/\/aipraktikum.com\/?p=15695"},"modified":"2026-06-05T18:30:39","modified_gmt":"2026-06-05T18:30:39","slug":"stari-promptovi-te-tiho-izdaju","status":"publish","type":"post","link":"https:\/\/aipraktikum.com\/en\/stari-promptovi-te-tiho-izdaju\/","title":{"rendered":"Stari promptovi te tiho izdaju"},"content":{"rendered":"
\n\t\t\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

Promptovi koji su savr\u0161eno radili lani na novim modelima poput GPT-5.5 i Claude Opus 4.7 sada daju lo\u0161ije rezultate. Razlog je jednostavan \u2014 novi modeli te shva\u0107aju doslovno<\/strong>. Tri navike koje mora\u0161 promijeniti: (1) prestani slijepo lijepiti \u201eti si ekspert” na po\u010detak, (2) eksplicitno reci modelu koje datoteke da pogleda, i (3) tra\u017ei od njega izvje\u0161taj kad odradi slo\u017een zadatak.<\/p>

Jesi li primijetio da ti AI odjednom \u201eglupi”, iako koristi\u0161 isti prompt kao prije?<\/p>

Nisi ti poludio. Promijenio se model, a ti nisi<\/strong> promijenio na\u010din kako mu govori\u0161.<\/p>

Evo \u0161to se dogodilo \u2014 i kako to popraviti danas.<\/p>

\u0160to se zapravo promijenilo<\/h2>

Stari modeli su naga\u0111ali. Ako bi im dao polovi\u010dnu uputu, popunili bi rupe sami \u2014 ponekad dobro, ponekad nasumi\u010dno. Bilo je to kao kolega koji te zna toliko dobro da dovr\u0161i re\u010denicu umjesto tebe.<\/p>

Novi modeli \u2014 GPT-5.5<\/strong> (iza\u0161ao u travnju 2026.) i Claude Opus 4.8<\/strong>\u00a0\u2014 rade druga\u010dije. Oni te slu\u0161aju doslovno<\/strong>. Svaka rije\u010d koju napi\u0161e\u0161 sada nosi ve\u0107u te\u017einu. To je odli\u010dna vijest ako si precizan. Lo\u0161a ako si navikao na nemar.<\/p>

Drugim rije\u010dima: AI je prestao \u010ditati tvoje misli i po\u010deo \u010ditati tvoje rije\u010di. A ve\u0107ina nas pi\u0161e promptove kao da AI jo\u0161 uvijek poga\u0111a.<\/p>

Ove tri navike te ko\u0161taju<\/p>

Gre\u0161ka #1: Lijepi\u0161 \u201eti si ekspert” na sve<\/h2>

Godinama smo u\u010dili: po\u010dni prompt s ulogom. \u201eTi si svjetski stru\u010dnjak za marketing.” \u201ePona\u0161aj se kao iskusni odvjetnik.” Postalo je refleks.<\/p>

Problem je \u0161to kod novih modela ta navika zna odmo\u0107i<\/strong>.<\/p>

\u0160to ka\u017ee istra\u017eivanje (ne naga\u0111anje)<\/h3>

Ovdje moram biti precizan, jer se na internetu o ovome pri\u010da pojednostavljeno. Studija istra\u017eiva\u010da s Carnegie Mellona, Stanforda i Sveu\u010dili\u0161ta u Michiganu (Zheng i suradnici, 2023.) testirala je 162 razli\u010dite uloge na 2.410 \u010dinjeni\u010dnih pitanja.<\/p>

Zaklju\u010dak: dodavanje persone ne pobolj\u0161ava<\/strong> to\u010dnost modela u odnosu na verziju bez uloge.<\/p>

Novije istra\u017eivanje sa Sveu\u010dili\u0161ta Ju\u017ene Kalifornije (USC, 2026.) dodaje klju\u010dnu nijansu \u2014 i ovo je dio koji ve\u0107ina ljudi presko\u010di:<\/p>

  • Kod zadataka vezanih uz stil i pisanje<\/strong> (ton, kreativni tekstovi, role-play) uloga poma\u017ee<\/strong>.<\/li>
  • Kod zadataka koji ovise o znanju<\/strong> (matematika, kod, \u010dinjenice) uloga daje lo\u0161ije<\/strong> rezultate.<\/li><\/ul>

    Teorija je da kad modelu ka\u017ee\u0161 \u201eti si ekspert”, on potro\u0161i dio pa\u017enje na to kako<\/em> zvu\u010dati kao ekspert umjesto da rije\u0161i sam zadatak. A re\u0107i mu da je stru\u010dnjak ne dodaje nijednu novu \u010dinjenicu \u2014 samo mu zamagli pristup onome \u0161to ve\u0107 zna.<\/p>

    Zna\u010di<\/h3>

    Nemoj slijepo slu\u0161ati savjet \u201eizbaci ulogu uvijek”. To je pola istine.<\/p>

    Pravilo koje (meni) radi:<\/strong> Ako pi\u0161e\u0161 tekst, oglas ili e-mail \u2014 uloga ti i dalje mo\u017ee pomo\u0107i oko tona. Ako tra\u017ei\u0161 analizu, brojke ili to\u010dne \u010dinjenice \u2014 izbaci ulogu i fokusiraj se na to \u0161to \u017eeli\u0161 i kako izgleda dobar rezultat<\/strong>.<\/p>

    Lo\u0161 prompt (stari refleks):<\/strong><\/p>

    \u201eTi si svjetski stru\u010dnjak za cijene. Pomozi mi poslo\u017eiti cjenovnu strategiju.”<\/p><\/blockquote>

    Pola prompta potro\u0161eno na ulogu. Pola na zadatak.<\/p>

    Bolji prompt (za novi model):<\/strong><\/p>

    \u201eDaj mi tri opcije za cjenovnu strategiju. Svaka neka vodi s jednim kompromisom (trade-off) i navedi izvor za svaki broj koji koristi\u0161.”<\/p><\/blockquote>

    Ku\u017ei\u0161\/vidi\u0161 razliku?<\/p>

    Drugi prompt ne tro\u0161i rije\u010di na glumu. Tro\u0161i ih na to da definira kako izgleda dobar rezultat. To je ono \u0161to novom modelu sada najvi\u0161e zna\u010di.<\/p>

    Gre\u0161ka #2: Pretpostavlja\u0161 da AI \u201ezna” gledati tvoje datoteke<\/h2>

    Ako koristi\u0161 Projects u Claudeu<\/strong> ili GPT projekte<\/strong>, ovo je za tebe.<\/p>

    Poanta projekata je da ima\u0161 AI usmjeren na jedan zadatak koji ponavlja\u0161 \u2014 i da u njega ubaci\u0161 kontekstne datoteke: tvoj brand voice, pro\u0161le ponude, cjenik, primjere dobrog rada.<\/p>

    Stari modeli bi ponekad sami pogodili da te datoteke trebaju i pogledali ih. Klju\u010dna rije\u010d: ponekad<\/strong>. I nasumi\u010dno<\/strong>.<\/p>

    Novi modeli ih ne\u0107e pogledati ako im to izri\u010dito ne ka\u017ee\u0161. Sjeti se \u2014 shva\u0107aju te doslovno. Ako nisi rekao \u201epogledaj datoteku”, za njih to zna\u010di \u201ene treba mi datoteka”.<\/p>

    Primjer iz prakse<\/h3>

    Zamisli vlasnika male agencije u Osijeku koji u Claude projektu ima brand voice, deset pro\u0161lih ponuda i cjenik. Tra\u017ei novu ponudu ovako:<\/p>

    Lo\u0161 prompt:<\/strong><\/p>

    \u201eNapi\u0161i ponudu za firmu X na temelju ovih bilje\u0161ki s discovery calla.”<\/p><\/blockquote>

    Nada se da \u0107e AI shvatiti da treba zaviriti u kontekstne datoteke. Stari model mo\u017eda bi. Novi ne\u0107e.<\/p>

    Bolji prompt:<\/strong><\/p>

    \u201eNapi\u0161i ponudu za firmu X. Pogledaj kontekstne datoteke u ovom projektu \u2014 konkretno: koristi brand voice<\/em> datoteku za ton, pro\u0161le ponude<\/em> za strukturu, i cjenik<\/em> za sve brojke u ponudi.”<\/p><\/blockquote>

    Mo\u017eda ti se \u010dini kao korak unatrag \u2014 \u201epa zar moram sve obja\u0161njavati?”. Nije korak unatrag. Stari model je naga\u0111ao koju datoteku otvoriti, \u0161to je nepouzdano. Novi model, kad mu jasno ka\u017ee\u0161, otvori je gotovo sigurno. Tro\u0161i\u0161 dvije re\u010denice vi\u0161e, a dobiva\u0161 pouzdanost.<\/p>

    Gre\u0161ka #3: Ne tra\u017ei\u0161 izvje\u0161taj kad zadatak postane velik<\/h2>

    Ovo je nova vrsta problema \u2014 pojavila se ba\u0161 zato \u0161to su modeli postali bolji.<\/p>

    Stari modeli su radili pet do deset koraka po zadatku. Lako provjeri\u0161 je li sve odra\u0111eno. Novi modeli mogu odraditi 50, 100 koraka, raditi pola sata pa i dulje \u2014 pogotovo u alatima poput Claude Cowork ili Codex na desktopu.<\/p>

    I tu nastaje zamka. Recimo da da\u0161 AI-ju 50 stvari za obraditi \u2014 50 transkripata, 50 faktura, 50 prezentacija. On vrati rezultat i ka\u017ee \u201egotovo”. Ali postoji \u0161ansa da je presko\u010dio osam<\/strong> od njih, a ti to nikad ne bi saznao bez ru\u010dne provjere svake stavke.<\/p>

    \u201eGotovo” vi\u0161e ne zna\u010di nu\u017eno \u201esve odra\u0111eno”. Zna\u010di \u201eja mislim da sam zavr\u0161io”.<\/p>

    Rje\u0161enje: natjeraj ga da broji<\/h3>

    Dodaj ovo na kraj svakog velikog zadatka:<\/p>

    \u201eKad zavr\u0161i\u0161, javi mi to\u010dan broj stavki koje si obradio \u2014 o\u010dekujem 50, pa o\u010dekujem da mi vrati\u0161 50. Za svaku stavku koju si presko\u010dio, imenuj je i reci za\u0161to. Rije\u010d ‘gotovo’ ili ‘zavr\u0161eno’ smije\u0161 koristiti tek kad si stvarno obradio sve stavke.”<\/p><\/blockquote>

    Tri stvari rade ovdje: tra\u017ei\u0161 broj (brza provjera), tra\u017ei\u0161 dokaz za svako preskakanje (s imenom i razlogom), i postavlja\u0161 pravilo da \u201egotovo” nije dozvoljeno dok nije stvarno gotovo. Jednostavno, ali radi.<\/p>

    Brzi podsjetnik \u2014 tri promjene<\/h2>
    Stara navika<\/th>\u0160to radi na novom modelu<\/th>\u0160to umjesto toga<\/th><\/tr><\/thead>
    \u201eTi si ekspert za\u2026” na sve<\/td>Tro\u0161i pa\u017enju na glumu, \u0161teti kod \u010dinjenica<\/td>Definiraj cilj i kako izgleda dobar rezultat (ulogu zadr\u017ei samo za ton\/pisanje)<\/td><\/tr>
    Pretpostavlja\u0161 da gleda datoteke<\/td>Ne gleda ih osim ako ne ka\u017ee\u0161<\/td>Imenuj to\u010dno koju datoteku za \u0161to<\/td><\/tr>
    Vjeruje\u0161 da \u201egotovo” zna\u010di gotovo<\/td>Mo\u017ee tiho presko\u010diti stavke<\/td>Tra\u017ei broj obra\u0111enih + popis presko\u010denog<\/td><\/tr><\/tbody><\/table>

    Jedno upozorenje na kraju<\/h2>

    Ova tri popravka ne\u0107e spasiti zadatak koji je jednostavno prevelik za chat prozor. Neke stvari spadaju na desktop, u alate s agentima (Claude Cowork, Codex), a ne u obi\u010dan chat. Ako stalno gu\u0161i\u0161 jedan razgovor s prevelikim zadatkom \u2014 problem nije prompt, nego mjesto gdje radi\u0161.<\/p>

    Ali za 90% svakodnevnog posla, ove tri promjene vrate ti kvalitetu koju si mislio da si izgubio.<\/p>

    AI se nije pokvario. Samo te po\u010deo slu\u0161ati doslovno. Vrijeme je da i ti njemu govori\u0161 jasnije.<\/p>

    \u00a0<\/p>

    \u00a0<\/p>

    FAQ<\/h2>

    Moram li uvijek izbaciti \u201eti si ekspert” iz prompta?<\/strong><\/p>

    Ne. Kod pisanja, oglasa i zadataka gdje je bitan ton \u2014 uloga ti i dalje mo\u017ee pomo\u0107i. Izbaci je kod zadataka koji tra\u017ee to\u010dne brojke, \u010dinjenice ili logiku, jer ondje zna odmo\u0107i.<\/p>

    Za\u0161to novi modeli ne gledaju moje datoteke sami od sebe?<\/strong><\/p>

    Jer te shva\u0107aju doslovno. Ako u promptu ne ka\u017ee\u0161 da trebaju pogledati odre\u0111enu datoteku, pretpostavljaju da im ne treba. Stari modeli su naga\u0111ali; novi slu\u0161aju upute.<\/p>

    \u0160to ako AI ka\u017ee da je gotov, a nije obradio sve?<\/strong><\/p>

    Dodaj u prompt zahtjev da ti vrati to\u010dan broj obra\u0111enih stavki i da imenuje svaku koju je presko\u010dio. Tako u sekundi vidi\u0161 je li broj koji si o\u010dekivao jednak broju koji si dobio.<\/p>

    Vrijedi li ovo samo za GPT-5.5 i Opus 4.7?<\/strong><\/p>

    Na\u010dela vrijede za sve novije modele koji bolje slijede upute. Konkretni nazivi se mijenjaju, ali smjer je isti: \u0161to su modeli bolji u slu\u0161anju, to su tvoje rije\u010di va\u017enije.<\/p>

    Je li ovo zna\u010di da je prompt engineering mrtav?<\/strong><\/p>

    Suprotno. Postaje va\u017eniji. Kad model prestane naga\u0111ati, jedino \u0161to ostaje je koliko jasno zna\u0161 re\u0107i \u0161to \u017eeli\u0161.<\/p>

    Gdje da po\u010dnem ako imam samo 10 minuta?

    <\/strong>Uzmi jedan prompt koji koristi\u0161 stalno. Izbaci nepotrebnu ulogu, dodaj re\u010denicu \u201eevo kako izgleda dobar rezultat”, i ako koristi\u0161 projekt \u2014 imenuj datoteke. To je to.<\/p>

    \u00a0<\/p>

    Darko Biland\u017ei\u0107
    <\/strong> info@aipraktikum.com
    U doba inteligencije, neznanje je odabir.<\/em><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<\/div>","protected":false},"excerpt":{"rendered":"

    Promptovi koji su savr\u0161eno radili lani na novim modelima poput GPT-5.5 i Claude Opus 4.7 sada daju lo\u0161ije rezultate. Razlog je jednostavan \u2014 novi modeli te shva\u0107aju doslovno. Tri navike koje mora\u0161 promijeniti: (1) prestani slijepo lijepiti \u201eti si ekspert” na…<\/p>","protected":false},"author":1,"featured_media":15710,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[],"class_list":["post-15695","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai"],"yoast_head":"\nStari promptovi te tiho izdaju - AI PRAKTIKUM<\/title>\n<meta name=\"description\" content=\"Stari promptovi daju lo\u0161ije rezultate na novim modelima. Tri konkretna popravka za GPT-5.5 i Claude Opus 4.7 \u2014 uz provjerene izvore\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/aipraktikum.com\/en\/stari-promptovi-te-tiho-izdaju\/\" \/>\n<meta property=\"og:locale\" content=\"en_GB\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Stari promptovi te tiho izdaju - AI PRAKTIKUM\" \/>\n<meta property=\"og:description\" content=\"Stari promptovi daju lo\u0161ije rezultate na novim modelima. Tri konkretna popravka za GPT-5.5 i Claude Opus 4.7 \u2014 uz provjerene izvore\" \/>\n<meta property=\"og:url\" content=\"https:\/\/aipraktikum.com\/en\/stari-promptovi-te-tiho-izdaju\/\" \/>\n<meta property=\"og:site_name\" content=\"AI PRAKTIKUM\" \/>\n<meta property=\"article:published_time\" content=\"2026-06-05T18:08:35+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2026-06-05T18:30:39+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/aipraktikum.com\/wp-content\/uploads\/2026\/06\/ChatGPT-Image-Jun-5-2026-08_28_03-PM.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"1672\" \/>\n\t<meta property=\"og:image:height\" content=\"941\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"author\" content=\"Darko Biland\u017ei\u0107\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Darko Biland\u017ei\u0107\" \/>\n\t<meta name=\"twitter:label2\" content=\"Estimated reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"6 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/aipraktikum.com\\\/stari-promptovi-te-tiho-izdaju\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/aipraktikum.com\\\/stari-promptovi-te-tiho-izdaju\\\/\"},\"author\":{\"name\":\"Darko Biland\u017ei\u0107\",\"@id\":\"https:\\\/\\\/aipraktikum.com\\\/#\\\/schema\\\/person\\\/34fc3671c6d5f23e856be6d5af423502\"},\"headline\":\"Stari promptovi te tiho izdaju\",\"datePublished\":\"2026-06-05T18:08:35+00:00\",\"dateModified\":\"2026-06-05T18:30:39+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/aipraktikum.com\\\/stari-promptovi-te-tiho-izdaju\\\/\"},\"wordCount\":1474,\"publisher\":{\"@id\":\"https:\\\/\\\/aipraktikum.com\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/aipraktikum.com\\\/stari-promptovi-te-tiho-izdaju\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/aipraktikum.com\\\/wp-content\\\/uploads\\\/2026\\\/06\\\/ChatGPT-Image-Jun-5-2026-08_28_03-PM.jpg\",\"articleSection\":[\"AI\"],\"inLanguage\":\"en-GB\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/aipraktikum.com\\\/stari-promptovi-te-tiho-izdaju\\\/\",\"url\":\"https:\\\/\\\/aipraktikum.com\\\/stari-promptovi-te-tiho-izdaju\\\/\",\"name\":\"Stari promptovi te tiho izdaju - AI PRAKTIKUM\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/aipraktikum.com\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/aipraktikum.com\\\/stari-promptovi-te-tiho-izdaju\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/aipraktikum.com\\\/stari-promptovi-te-tiho-izdaju\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/aipraktikum.com\\\/wp-content\\\/uploads\\\/2026\\\/06\\\/ChatGPT-Image-Jun-5-2026-08_28_03-PM.jpg\",\"datePublished\":\"2026-06-05T18:08:35+00:00\",\"dateModified\":\"2026-06-05T18:30:39+00:00\",\"description\":\"Stari promptovi daju lo\u0161ije rezultate na novim modelima. Tri konkretna popravka za GPT-5.5 i Claude Opus 4.7 \u2014 uz provjerene izvore\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/aipraktikum.com\\\/stari-promptovi-te-tiho-izdaju\\\/#breadcrumb\"},\"inLanguage\":\"en-GB\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/aipraktikum.com\\\/stari-promptovi-te-tiho-izdaju\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-GB\",\"@id\":\"https:\\\/\\\/aipraktikum.com\\\/stari-promptovi-te-tiho-izdaju\\\/#primaryimage\",\"url\":\"https:\\\/\\\/aipraktikum.com\\\/wp-content\\\/uploads\\\/2026\\\/06\\\/ChatGPT-Image-Jun-5-2026-08_28_03-PM.jpg\",\"contentUrl\":\"https:\\\/\\\/aipraktikum.com\\\/wp-content\\\/uploads\\\/2026\\\/06\\\/ChatGPT-Image-Jun-5-2026-08_28_03-PM.jpg\",\"width\":1672,\"height\":941,\"caption\":\"Novi modeli te shva\u0107aju doslovno \u2014 tvoje rije\u010di sada nose ve\u0107u te\u017einu.\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/aipraktikum.com\\\/stari-promptovi-te-tiho-izdaju\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"AI\",\"item\":\"https:\\\/\\\/aipraktikum.com\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Stari promptovi te tiho izdaju\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/aipraktikum.com\\\/#website\",\"url\":\"https:\\\/\\\/aipraktikum.com\\\/\",\"name\":\"AI PRAKTIKUM\",\"description\":\"Umjetna inteligencija za poduzetnike\",\"publisher\":{\"@id\":\"https:\\\/\\\/aipraktikum.com\\\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/aipraktikum.com\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-GB\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/aipraktikum.com\\\/#organization\",\"name\":\"AI PRAKTIKUM\",\"url\":\"https:\\\/\\\/aipraktikum.com\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-GB\",\"@id\":\"https:\\\/\\\/aipraktikum.com\\\/#\\\/schema\\\/logo\\\/image\\\/\",\"url\":\"https:\\\/\\\/aipraktikum.com\\\/2026\\\/wp-content\\\/uploads\\\/2025\\\/10\\\/cropped-AI-PRAKTIKUM-512-x-512-px.png\",\"contentUrl\":\"https:\\\/\\\/aipraktikum.com\\\/2026\\\/wp-content\\\/uploads\\\/2025\\\/10\\\/cropped-AI-PRAKTIKUM-512-x-512-px.png\",\"width\":512,\"height\":512,\"caption\":\"AI PRAKTIKUM\"},\"image\":{\"@id\":\"https:\\\/\\\/aipraktikum.com\\\/#\\\/schema\\\/logo\\\/image\\\/\"},\"sameAs\":[\"https:\\\/\\\/www.tiktok.com\\\/@darkobilandzic\"]},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/aipraktikum.com\\\/#\\\/schema\\\/person\\\/34fc3671c6d5f23e856be6d5af423502\",\"name\":\"Darko Biland\u017ei\u0107\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-GB\",\"@id\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/dc648c5e86f8a9a7978578db75069ce0b212dbf68ccc0583df14bab7add81146?s=96&d=mm&r=g\",\"url\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/dc648c5e86f8a9a7978578db75069ce0b212dbf68ccc0583df14bab7add81146?s=96&d=mm&r=g\",\"contentUrl\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/dc648c5e86f8a9a7978578db75069ce0b212dbf68ccc0583df14bab7add81146?s=96&d=mm&r=g\",\"caption\":\"Darko Biland\u017ei\u0107\"},\"sameAs\":[\"https:\\\/\\\/aipraktikum.com\\\/\"],\"url\":\"https:\\\/\\\/aipraktikum.com\\\/en\\\/author\\\/creatos\\\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Stari promptovi te tiho izdaju - AI PRAKTIKUM","description":"Stari promptovi daju lo\u0161ije rezultate na novim modelima. Tri konkretna popravka za GPT-5.5 i Claude Opus 4.7 \u2014 uz provjerene izvore","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/aipraktikum.com\/en\/stari-promptovi-te-tiho-izdaju\/","og_locale":"en_GB","og_type":"article","og_title":"Stari promptovi te tiho izdaju - AI PRAKTIKUM","og_description":"Stari promptovi daju lo\u0161ije rezultate na novim modelima. Tri konkretna popravka za GPT-5.5 i Claude Opus 4.7 \u2014 uz provjerene izvore","og_url":"https:\/\/aipraktikum.com\/en\/stari-promptovi-te-tiho-izdaju\/","og_site_name":"AI PRAKTIKUM","article_published_time":"2026-06-05T18:08:35+00:00","article_modified_time":"2026-06-05T18:30:39+00:00","og_image":[{"width":1672,"height":941,"url":"https:\/\/aipraktikum.com\/wp-content\/uploads\/2026\/06\/ChatGPT-Image-Jun-5-2026-08_28_03-PM.jpg","type":"image\/jpeg"}],"author":"Darko Biland\u017ei\u0107","twitter_card":"summary_large_image","twitter_misc":{"Written by":"Darko Biland\u017ei\u0107","Estimated reading time":"6 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/aipraktikum.com\/stari-promptovi-te-tiho-izdaju\/#article","isPartOf":{"@id":"https:\/\/aipraktikum.com\/stari-promptovi-te-tiho-izdaju\/"},"author":{"name":"Darko Biland\u017ei\u0107","@id":"https:\/\/aipraktikum.com\/#\/schema\/person\/34fc3671c6d5f23e856be6d5af423502"},"headline":"Stari promptovi te tiho izdaju","datePublished":"2026-06-05T18:08:35+00:00","dateModified":"2026-06-05T18:30:39+00:00","mainEntityOfPage":{"@id":"https:\/\/aipraktikum.com\/stari-promptovi-te-tiho-izdaju\/"},"wordCount":1474,"publisher":{"@id":"https:\/\/aipraktikum.com\/#organization"},"image":{"@id":"https:\/\/aipraktikum.com\/stari-promptovi-te-tiho-izdaju\/#primaryimage"},"thumbnailUrl":"https:\/\/aipraktikum.com\/wp-content\/uploads\/2026\/06\/ChatGPT-Image-Jun-5-2026-08_28_03-PM.jpg","articleSection":["AI"],"inLanguage":"en-GB"},{"@type":"WebPage","@id":"https:\/\/aipraktikum.com\/stari-promptovi-te-tiho-izdaju\/","url":"https:\/\/aipraktikum.com\/stari-promptovi-te-tiho-izdaju\/","name":"Stari promptovi te tiho izdaju - AI PRAKTIKUM","isPartOf":{"@id":"https:\/\/aipraktikum.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/aipraktikum.com\/stari-promptovi-te-tiho-izdaju\/#primaryimage"},"image":{"@id":"https:\/\/aipraktikum.com\/stari-promptovi-te-tiho-izdaju\/#primaryimage"},"thumbnailUrl":"https:\/\/aipraktikum.com\/wp-content\/uploads\/2026\/06\/ChatGPT-Image-Jun-5-2026-08_28_03-PM.jpg","datePublished":"2026-06-05T18:08:35+00:00","dateModified":"2026-06-05T18:30:39+00:00","description":"Stari promptovi daju lo\u0161ije rezultate na novim modelima. Tri konkretna popravka za GPT-5.5 i Claude Opus 4.7 \u2014 uz provjerene izvore","breadcrumb":{"@id":"https:\/\/aipraktikum.com\/stari-promptovi-te-tiho-izdaju\/#breadcrumb"},"inLanguage":"en-GB","potentialAction":[{"@type":"ReadAction","target":["https:\/\/aipraktikum.com\/stari-promptovi-te-tiho-izdaju\/"]}]},{"@type":"ImageObject","inLanguage":"en-GB","@id":"https:\/\/aipraktikum.com\/stari-promptovi-te-tiho-izdaju\/#primaryimage","url":"https:\/\/aipraktikum.com\/wp-content\/uploads\/2026\/06\/ChatGPT-Image-Jun-5-2026-08_28_03-PM.jpg","contentUrl":"https:\/\/aipraktikum.com\/wp-content\/uploads\/2026\/06\/ChatGPT-Image-Jun-5-2026-08_28_03-PM.jpg","width":1672,"height":941,"caption":"Novi modeli te shva\u0107aju doslovno \u2014 tvoje rije\u010di sada nose ve\u0107u te\u017einu."},{"@type":"BreadcrumbList","@id":"https:\/\/aipraktikum.com\/stari-promptovi-te-tiho-izdaju\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"AI","item":"https:\/\/aipraktikum.com\/"},{"@type":"ListItem","position":2,"name":"Stari promptovi te tiho izdaju"}]},{"@type":"WebSite","@id":"https:\/\/aipraktikum.com\/#website","url":"https:\/\/aipraktikum.com\/","name":"AI PRAKTIKUM","description":"Umjetna inteligencija za poduzetnike","publisher":{"@id":"https:\/\/aipraktikum.com\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/aipraktikum.com\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-GB"},{"@type":"Organization","@id":"https:\/\/aipraktikum.com\/#organization","name":"AI PRAKTIKUM","url":"https:\/\/aipraktikum.com\/","logo":{"@type":"ImageObject","inLanguage":"en-GB","@id":"https:\/\/aipraktikum.com\/#\/schema\/logo\/image\/","url":"https:\/\/aipraktikum.com\/2026\/wp-content\/uploads\/2025\/10\/cropped-AI-PRAKTIKUM-512-x-512-px.png","contentUrl":"https:\/\/aipraktikum.com\/2026\/wp-content\/uploads\/2025\/10\/cropped-AI-PRAKTIKUM-512-x-512-px.png","width":512,"height":512,"caption":"AI PRAKTIKUM"},"image":{"@id":"https:\/\/aipraktikum.com\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/www.tiktok.com\/@darkobilandzic"]},{"@type":"Person","@id":"https:\/\/aipraktikum.com\/#\/schema\/person\/34fc3671c6d5f23e856be6d5af423502","name":"Darko Biland\u017ei\u0107","image":{"@type":"ImageObject","inLanguage":"en-GB","@id":"https:\/\/secure.gravatar.com\/avatar\/dc648c5e86f8a9a7978578db75069ce0b212dbf68ccc0583df14bab7add81146?s=96&d=mm&r=g","url":"https:\/\/secure.gravatar.com\/avatar\/dc648c5e86f8a9a7978578db75069ce0b212dbf68ccc0583df14bab7add81146?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/dc648c5e86f8a9a7978578db75069ce0b212dbf68ccc0583df14bab7add81146?s=96&d=mm&r=g","caption":"Darko Biland\u017ei\u0107"},"sameAs":["https:\/\/aipraktikum.com\/"],"url":"https:\/\/aipraktikum.com\/en\/author\/creatos\/"}]}},"_links":{"self":[{"href":"https:\/\/aipraktikum.com\/en\/wp-json\/wp\/v2\/posts\/15695","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/aipraktikum.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/aipraktikum.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/aipraktikum.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/aipraktikum.com\/en\/wp-json\/wp\/v2\/comments?post=15695"}],"version-history":[{"count":13,"href":"https:\/\/aipraktikum.com\/en\/wp-json\/wp\/v2\/posts\/15695\/revisions"}],"predecessor-version":[{"id":15709,"href":"https:\/\/aipraktikum.com\/en\/wp-json\/wp\/v2\/posts\/15695\/revisions\/15709"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/aipraktikum.com\/en\/wp-json\/wp\/v2\/media\/15710"}],"wp:attachment":[{"href":"https:\/\/aipraktikum.com\/en\/wp-json\/wp\/v2\/media?parent=15695"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/aipraktikum.com\/en\/wp-json\/wp\/v2\/categories?post=15695"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/aipraktikum.com\/en\/wp-json\/wp\/v2\/tags?post=15695"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}