Claude aur ChatGPT 101: Aik Crash Course
9 concepts. 2 cockpits. 1 discipline: un do AI assistants ke andar aitmaad se kaam karne ka tareeqa jinhein aap sab se zyada istemaal karein ge.
Aap AI ke saath kaam karne ke bunyadi ideas pehle se jaante hain. Aap jaante hain ke model ko context chahiye. Aap jaante hain ke task ko kaise bayan karna hai. Aap jaante hain ke kuch kaam delegate karne chahiye aur kuch apne paas rakhne chahiye.
Phir aap Claude ya ChatGPT kholte hain aur aik mukhtalif masla saamne aata hai: sab kuch hai kahan?
Model picker hai. Plus menu hai. Projects hain. Memory hai. Skills hain. Plugins hain. Connectors ya apps hain. Search hai. Research hai. Thinking modes hain.
Ideas pehchane hue hain. Controls nahin.
Yeh course isi masle ko hal karta hai.
Aap Claude aur ChatGPT ko saath saath seekhein ge. Maqsad har button yaad karna nahin hai. Buttons ki jagah badalti hai. Product names badalte hain. Plans badalte hain.
Maqsad interface ke neeche maujood pattern seekhna hai.
Projects kaam ko aik jagah rakhte hain. Memory aap ke bare mein kaam ka context sambhalti hai. Skills baar baar istemaal hone wale kaam ke tareeqe package karti hain. Connectors aur apps assistant ko external systems tak pohnch dete hain. Research modes aap ke liye information jama karte aur us ka mutalea karte hain.
Product badalne par bhi discipline wahi rehta hai.
Isi liye yeh course aik nahin, do cockpits sikhata hai.
Parhne ka waqt: taqreeban 30 se 35 minute, aur practice prompts aur self-check ke liye mazeed taqreeban 25 minute.
Software tezi se badalta hai. Is page par har product claim ko 25 August 2026 ko Anthropic aur OpenAI ki public documentation se verify kiya gaya tha.
Agar aap ke parhne tak koi button jagah badal chuka ho, to pareshan na hon. Is course ka mental model yaad rakhein, phir support.claude.com aur help.openai.com par mojooda help pages dekhein.
Plan availability features se bhi zyada tezi se badalti hai, is liye un par bharosa karne se pehle current pricing aur plan limits confirm karein.
Yeh Foundations ka chautha course hai. Yeh maan kar chalta hai ke aap pehle hi yeh parh chuke hain:
Yeh course un ideas ko dobara nahin sikhata. Yeh dikhata hai ke woh Claude aur ChatGPT ke andar kahan milte hain.
| Topic | Kahan sikhaya gaya | Yeh course kya karta hai |
|---|---|---|
| Models aur context windows kaise kaam karte hain | AI Asal Mein Kya Hai | Un ideas ko istemaal karta hai |
| 4D framework | AI Fluency | Use product features par apply karta hai |
| Prompting technique | 2026 Mein AI Prompting | Use asli workflows mein istemaal karta hai |
| Concepts ke taur par skills aur MCP | Skills aur Connectors | Controls ka taaruf karata hai |
| Desktop apps, coding tools, agentic hand-off | Professionals ke Liye Cowork aur OpenWork aur General Agents courses | Sirf chhota sa orientation deta hai |
| Claude aur ChatGPT ke chat workspaces | Yeh course | Unhein seedha sikhata hai |
Mukhtasar yeh ke: yeh course chat workspace ke andar rehta hai. Desktop agents, coding environments, aur bare agentic systems baad mein aate hain.
Do Minute Mein Dekhein
Tour shuru karne se pehle cockpit ko apna taaruf khud karne dein.
Claude kholein. Yeh paste karein aur jawab parhein:
List every control I can see in this workspace right now, and tell me
in one line what each one is for. Just the controls, no advice.
Ab doosre tab mein ChatGPT kholein aur bilkul yahi cheez paste karein.
Dono lists ko saath saath rakh dein.
Aap zyada tar unhi saat cheezon ke liye do mukhtalif vocabularies dekhein ge. Aik artifact kehta hai, doosra writing block. Aik connector kehta hai, doosra app. Aik project knowledge kehta hai, doosra project files. Dono model, memory, instructions, search, aur research kehte hain.
Yahi poora course aik screenshot mein hai. Controls ke naam mukhtalif hain. Aap un se jo kaam karte hain woh aik hi hai.
Aik imaandaar warning hai, aur kaam ki warning hai. Assistant apne interface ke bare mein hamesha bharosemand nahin hota, is liye list mein koi aisi cheez shamil ho sakti hai jo asal mein wahan na ho, ya koi mojood cheez reh sakti hai. Is baat ko note karein. Kisi pur-aitmaad jawab ko khud cheez se mila kar dekhna Concept 9 hai, aur aap course ke do minute mein hi us se mil chuke hain.
Woh Sawal Jin Ka Yeh Course Jawab Deta Hai
Aakhir tak aapko in sab ke jawab dene aane chahiye.
Hissa 1: Cockpit
| # | Sawal | Kahan jawab diya gaya |
|---|---|---|
| 1 | Sirf aik chunne ke bajaye Claude aur ChatGPT dono kyun seekhein? | Concept 1 |
| 2 | Dono mein kya aik jaisa hai aur kya mukhtalif? | Concept 1 |
| 3 | Jab model names badalte rehte hain to model kaise chunoon? | Concept 2 |
| 4 | Thinking mode kab istemaal karna chahiye? | Concept 2 |
| 5 | File upload karne par asal mein kya hota hai? | Concept 3 |
Hissa 2: Workspace Ko Apna Banana
| # | Sawal | Kahan jawab diya gaya |
|---|---|---|
| 6 | Project kya hai aur mujhe kab banana chahiye? | Concept 4 |
| 7 | Instructions, memory, aur projects mein kya farq hai? | Concept 5 |
| 8 | Artifacts aur writing blocks kya hain? | Concept 6 |
| 9 | Skills aur plugins kya hain, aur yeh projects aur custom GPTs se kaise mukhtalif hain? | Concept 7 |
Hissa 3: Rasai Barhana
| # | Sawal | Kahan jawab diya gaya |
|---|---|---|
| 10 | Connector ya app assistant ko kis cheez tak pohnchne deta hai? | Concept 8 |
| 11 | Search, thinking, ya research kab istemaal karna chahiye? | Concept 8 |
| 12 | Mujhe kaise pata chale ke assistant mere kaam mein waqai acha hai? | Concept 9 |
| 13 | Purane kaam par testing kya sabit karti hai, aur kya sabit nahin karti? | Concept 9 |
Hissa 1: Cockpit
1. Do Cockpits, Aik Discipline
Claude aur ChatGPT ko saath saath kholein. Dono hairat-angez had tak aik jaise dikhte hain.
Dono aapko yeh dete hain:
- pichhli chats ki list
- message box
- model ya reasoning control
- files aur tools ka menu
- projects ya workspaces
- memory aur instructions
- search aur research features
Yeh milta-julta hona aham hai.
Aapko kaam karne ke do bilkul mukhtalif tareeqe seekhne ki zaroorat nahin. Aapko aik discipline seekhni hai aur phir dekhna hai ke har product ne controls kahan rakhe hain.
Yeh skills aik product se doosre mein jaati hain:
- task ko saaf taur par bayan karna
- kaam ka context dena
- yeh chunna ke kya delegate karna hai
- jawab check karna
- yeh jaanna ke mazboot model kab istemaal karna hai
- sensitive information ki hifazat karna
Products ke names, layout, aur strengths mukhtalif hain. Yeh farq waqt ke saath badalte hain.
Bunyadi kaam nahin badalta.
Cockpit ke muqable par aik lamhe ke liye imaandari zaroori hai, kyun ke yeh ghalat tasawwur de sakta hai. Pilots do aircraft ko aik doosre ka badal nahin samajhte. Woh aik baar urhna seekhte hain, phir kuch waqt laga kar samajhte hain ke kisi khaas aircraft ke switches kahan hain. Yahan bhi yahi durust tawaqqo hai. Urhna woh hissa hai jo transfer hota hai. Switch layout woh hissa hai jise aap dekhte hain, aur update ke baad dobara dekhte hain.
Dono products jaanna aik amli wajah se bhi kaam aata hai. Claude aur ChatGPT har task par aik jaisi performance nahin dete. Jo model aik qisam ke kaam mein behtareen ho, woh doosri qisam mein kamzor ho sakta hai.
Aham kaam ke liye aik achi aadat yeh hai ke wahi task dono mein chalayein aur muqabla karein.
Yahi platform awareness hai: aik interface ke wafadar banne ke bajaye yeh jaanna ke har tool kis kaam mein acha hai.
Orientation ki aik aur baat. Dono companies ab chat ke ilawa bahut kuch deti hain: desktop apps, coding tools, browser agents, aur doosre agentic products. Woh is book mein baad mein aate hain.
Filhaal chat window ke andar rahen.
Beginner Ki Sab Se Aham Aadat
Assistant se usi tarah baat karein jaise kisi qabil colleague se karte hain.
Koi sanjeeda request bhejne se pehle teen cheezein cover karein:
- Mahaul wazeh karein. Aap kaun hain aur kya hasil karna chahte hain?
- Task define karein. Assistant ko bilkul kya karna chahiye?
- Rules set karein. Woh kis tone, format, constraints, ya examples par chale?
Interface achi description ki jagah nahin leta. Woh aapki description ko rehne ki jagah deta hai.
Yaad rakhein: button ki jagah nahin, pattern seekhein.
2. Models Aur Thinking Modes
Claude aur ChatGPT dono aapko chunne dete hain ke kisi task par kitni capability kharch karni hai.
Beginners aam taur par model names par tawajjoh dete hain. Yeh yaad rakhne ki ghalat cheez hai, kyun ke names tezi se badalte hain.
Is ke bajaye teen-level pattern seekhein.
| Level | Kis ke liye behtareen | Trade-off |
|---|---|---|
| Fast default | Rozmarra ke sawal, drafting, summaries, routine kaam | Tez aur sasta, mushkil maslon par kam gehra |
| Thinking ya reasoning | Multi-step reasoning, analysis, math, mushkil code | Dheema, zyada ehtiyat wala |
| Heavy flagship | Sab se mushkil analysis aur lambe, demanding tasks | Sab se dheema aur mehnga, sab se zyada qabil |
Claude aur ChatGPT in controls ko mukhtalif tareeqe se dikhate hain. Labels mein Sonnet, Opus, Instant, Thinking, Pro, ya koi version number jaise names ho sakte hain.
Apna mental model labels ke gird na banayein.
Use trade-off ke gird banayein:
pehle tez, phir zyada reasoning, phir maximum capability
Do Saade Rules
Rule 1: Tez se shuru karein.
Zyada tar rozmarra ke kaam ke liye normal fast mode istemaal karein.
Chhoti rewrite, summary, ya seedhi explanation ke liye heavy reasoning ka intizaar na karein.
Rule 2: Task mushkil ho to level barhayein.
Thinking mode ya zyada mazboot model tab istemaal karein jab kaam mein yeh cheezein hon:
- logic ke kai steps
- ehtiyat se muqabla
- riyaazi
- mushkil debugging
- aisa code jo durust hona lazmi ho
- pechida analysis
Kabhi kabhi "AI nakaam ho gaya" ka asal matlab hota hai "mein ne capability ka ghalat level istemaal kiya."
Model ko task ke mutabiq chunein, aadat ke mutabiq nahin.
- Rozmarra ke kaam ke liye Fast
- Mushkil reasoning ke liye Thinking
- Waqai mushkil kaam ke liye Flagship
Names badalne par bhi pattern kaam karta hai.
Yaad rakhein: pehle reasoning ka level chunein, phir current model names seekhein.
3. Woh Context Jo Aap Attach Karte Hain
Language model sirf us information ke saath kaam kar sakta hai jo us ke saamne ho.
Isi liye attachments aham hain.
Claude aur ChatGPT dono aapko yeh cheezein upload karne dete hain:
- PDFs
- Word documents
- spreadsheets
- CSV files
- images
- screenshots
- code files
File attach karne par aap assistant ko us conversation ke liye mazeed context dete hain.
In dono requests ka muqabla karein:
Is contract ka khulasa karein.
aur
Is contract ka khulasa karein. (contract attached hai)
Alfaaz taqreeban aik jaise hain. Doosri request kaam ki hai kyun ke assistant asal contract dekh sakta hai.
Screenshots par bhi yahi baat lagu hoti hai. Agar aap dashboard, error message, chart, ya interface mein madad chahte hain, to use dikhana aam taur par yaad se bayan karne se behtar hota hai.
Dono products live web bhi search kar sakte hain. Jab model ka built-in knowledge kaafi na ho, search current information ko conversation mein le aati hai.
Aur dono current conversation ko yaad rakhte hain, is liye pehle kahi hui baat baad ke jawabon ko shape karti rehti hai.
Woh Aadat Jo Banani Hai
Aham sawal poochne se pehle khud se poochein:
Kisi qabil insani colleague ko acha jawab dene se pehle kya dekhna zaroori hota?
Phir assistant ko woh material dein.
Yaad rakhein: zyada behtar context aam taur par clever prompt se zyada jawab sudharta hai.
Hissa 2: Workspace Ko Apna Banana
4. Projects: Kaam Ke Aik Silsile Ke Liye Kamra
AI istemaal karte hue kuch hafton baad aapki sidebar ulajh jaati hai.
Aap ke paas hoti hain:
- aik client ke bare mein aik chat
- usi client ke bare mein doosri chat
- teesri chat jis mein kaam ki file hai
- chauthi chat jahan aap ne wahi background phir se samjhaya
Projects isi masle ko hal karte hain.
Aik project kaam ke aik silsile ka workspace hota hai.
Use aik aisa kamra samjhein jahan aik topic se mutalliq har cheez aik jagah rehti hai.
Aik project aam taur par teen cheezein rakhta hai:
- Us kaam se mutalliq Chats
- Knowledge, yaani us project ke liye upload ki hui files
- Instructions, yaani project ke andar assistant ke kaam ke liye hamesha lagu rehne wali guidance
Maan lein aap Quarterly Board Reporting naam ka project banate hain.
Us ke andar aap yeh rakh sakte hain:
- pichhli quarter ki board deck
- current financial model
- aapki reporting style guide
- tone aur audience ke bare mein instructions
- board report se mutalliq har conversation
Ab aapko wahi files dobara upload ya wahi instructions baar baar dene ki zaroorat nahin.
Yahi poora maqsad hai: project kaam ke aik silsile ke context ko mustaqil bana deta hai.
Is book ki zaban mein, project aik chhota System of Context hai.
Claude Aur ChatGPT Mein Thora Farq Hai
Claude barhti hui knowledge ko retrieval se sambhalta hai. Jab project knowledge context window mein seedha fit hone se zyada ho jaati hai, to Claude sab kuch load karne ke bajaye us knowledge mein search karna shuru karta hai. Anthropic ke mutabiq is se capacity das guna tak barh sakti hai. Yeh switch khud hota hai aur paid-plan feature hai.
ChatGPT project memory shamil karta hai. Project do mein se aik memory istemaal kar sakta hai:
- default memory, jahan aapki wasi memory bhi hissa le sakti hai
- project-only memory, jahan project waqai boundary ban jata hai aur bahar ki memories andar nahin aatin
Is liye project hamesha organization mein madad karta hai. Project-only memory use isolation mein badalti hai.
Project Kab Banana Chahiye?
Yeh test istemaal karein:
Agar aap ne wahi background samjhaya ya wahi file teen baar upload ki hai, to us kaam ka project banna chahiye.
Aik baar ke sawal ko project ki zaroorat nahin.
Yaad rakhein: projects jaari rehne wale kaam ke liye hain.
5. Memory Aur Standing Instructions
Teen features context ko mustaqil banate hain:
- standing instructions
- memory
- projects
Beginners inhein aksar aik doosre mein mila dete hain. Saaf farq yeh hai.
Standing Instructions: Stable Rules
Standing instructions woh rules hain jo aap aik baar likhte hain aur chahte hain ke woh wasi taur par lagu hon.
Misaal ke taur par:
- "Explanation se pehle jawab dein."
- "Jab tak mein technical detail na maangoon, saadi zaban istemaal karein."
- "Mein software engineers ke liye nahin, business leaders ke liye likhta hun."
Claude is qisam ki preference ko settings aur styles mein rakhta hai. ChatGPT ise Settings aur Personalization ke neeche Custom Instructions mein rakhta hai.
Standing instructions un cheezon ke liye istemaal karein jo stable hon.
Memory: Aap Ke Bare Mein Evolving Context
Memory woh cheez hai jo assistant aapki conversations se seekhta hai aur baad mein istemaal kar sakta hai.
Is mein yeh shamil ho sakta hai:
- aapki batayi hui preferences
- baar baar aane wale goals
- aap ke kaam karne ke tareeqe ke facts
- koi bhi cheez jo aap ne use khaas taur par yaad rakhne ko kahi
Dono products aapko memory review, edit, aur delete karne dete hain. Work account par pehle check karein ke yeh on bhi hai ya nahin: Claude individual plans par memory default taur par on karta hai, aur Team aur Enterprise par tab tak off rakhta hai jab tak owner use enable na kare.
Dono aik aisa mode bhi dete hain jo conversation ko normal history aur memory se bahar rakhta hai: Claude ise incognito chat kehta hai, ChatGPT temporary chat. Dono vendors safety aur abuse review ke liye limited waqt, taqreeban tees din, aik copy phir bhi rakhte hain. Is liye in modes ko memory-free samjhein, trace-free nahin.
Projects: Kaam Ke Aik Silsile Ka Context
Projects dono se mukhtalif hain.
Project aik khaas client, course, research topic, product, ya jaari kaam ke kisi aur silsile ka context rakhta hai.
Seedha Rule
Stable rules ke liye instructions. Evolving context ke liye memory. Scoped work ke liye projects.
Yeh aik rule zyada tar configuration mistakes se bachata hai.
Yahan sirf asaani ka sawal nahin. Stored context mein sensitive ya private information ho sakti hai.
Jo yaad rakha gaya hai use review karein. Jo nahin rehna chahiye use delete karein. Jab mozo ka taqaza ho to incognito ya temporary chats istemaal karein. Work account par apni organization ki policy par chalein.
Memory kaam ki hai kyun ke context waqt ke saath jama ho kar zyada faida deta hai. Yeh woh jagah bhi hai jahan ehtiyat zaroori hai.
Achi tarah configured assistant aksar kuch hafton ke istemaal ke baad zyada kaam ka lagta hai. Model nahin badla. Us ke gird context behtar hua.
Yaad rakhein: persistence tabhi madad karti hai jab sahi information sahi jagah ho.
6. Artifacts Aur Writing Blocks: Woh Kaam Jo Aap Saath Le Ja Sakein
Chat conversation ke liye achi hai.
Mukammal kaam rakhne ke liye buri container hai.
Agar aap 2,000 alfaaz ki report, kaam karta calculator, diagram, ya chhoti application maangte hain, to aap chahte hain ke nateeja scroll mein aik aur message nahin, balke alag object ho.
Claude aur ChatGPT is masle ko mukhtalif tareeqe se hal karte hain.
Claude: Artifacts
Claude artifacts istemaal karta hai.
Artifact alag output hota hai jo conversation ke saath khulta hai. Woh yeh ho sakta hai:
- document
- code
- web page
- diagram
- vector image
- calculator
- dashboard
- chhoti interactive application
Aap Claude se baat jaari rakhte hain jab ke artifact wahi us cheez ke taur par rehta hai jo ban aur revise ho rahi hai.
Claude ke paas Word, Excel, PowerPoint, aur PDF files banane ki alag abilities bhi hain.
ChatGPT: Writing Blocks Aur Code Blocks
Purane ChatGPT tutorials canvas ka zikr karte hain, jo side-by-side editing panel tha.
Woh baat purani ho chuki hai. 2026 mein OpenAI ne current models par canvas retire kar diya aur wahi kaam conversation ke andar writing blocks aur code blocks mein le aya: editable regions jo thread ke andar hi rehte hain. Agar koi tutorial aapko canvas kholne ko kehta hai aur woh mil nahin raha, to wajah yahi hai.
Design mukhtalif hai. Idea aik hi hai:
kaam ki cheez ko us ke bare mein discussion se alag rakhein.
Behtar Results Kaise Milein
Sirf "dashboard" ya "report" na maangein.
Bayan karein ke mukammal cheez ko kya karna chahiye.
Misaal ke taur par:
Monthly budget tracker banayein. Mein category ke hisaab se expenses enter kar sakoon, pie chart dekh sakoon, aur budget se bahar jane par warning mile.
Yeh is se behtar hai:
Budget tracker banayein.
Yeh bhi batayein ke woh kis ke liye hai. Naye employees ka flowchart tajurbekar engineers ke flowchart jaisa object nahin hota.
Phir aik waqt mein aik change revise karein.
Yahi woh concept hai jahan bahut se beginners ko pehli asli hairat hoti hai: aap kaam ka software bayan karte hain aur kaam karta software saamne aa jata hai.
Baad ke do courses ise theek tarah khol kar samjhate hain:
Yaad rakhein: chat conversation hai; artifacts, writing blocks, aur code blocks kaam ko rakhte hain.
7. Skills Aur Plugins: Kaam Ke Packaged Tareeqe
Projects aik sawal ka jawab dete hain:
Mein sahi knowledge ko aik jagah kaise rakhoon?
Baar baar hone wala kaam mukhtalif sawal uthata hai:
Mein assistant se har baar wahi tareeqa kaise follow karwaoon?
Skills isi kaam ke liye hain.
Skill Kya Hai?
Skill kisi task ko karne ka packaged tareeqa hoti hai.
Is mein yeh ho sakte hain:
- instructions
- examples
- supporting resources
- kabhi kabhi code
Jab milta-julta task saamne aaye to assistant skill load karta hai.
Aap in kaamon ke liye skill bana sakte hain:
- quarterly business reviews likhna
- checklist ke muqable mein contract check karna
- customer meeting brief tayyar karna
- raw notes ko apne pasandeeda report format mein badalna
Har chat mein tareeqa dobara sikhane ke bajaye aap use aik baar package karte hain.
Claude ke paas skills hain. ChatGPT ke paas bhi skills hain.
Products mein farq hai ke skills kahan available aur kaise manage hoti hain, lekin bunyadi idea ab dono mein mushtarik hai.
Yeh Aham Kyun Hai
Dono vendors Agent Skills open standard par bante hain, jo agentskills.io par publish hua hai.
Skill Markdown files ka folder hoti hai jis ke peeche koi server ya runtime nahin hota, aur isi wajah se woh aik jagah se doosri jagah ja sakti hai. Kai companies ke darjanon tools ab wahi format parhte hain, is liye aik product ke liye likhi skill doosre mein install ho sakti hai.
Is se workflow kisi aik vendor mein phanse prompt se zyada paaedaar ban jata hai.
Yahi woh lesson bhi hai jis se yeh course shuru hua:
discipline mustaqil hai; tool badalne wali cheez hai.
Plugin Kya Hai?
Plugin kisi wasi qisam ke kaam ke liye capabilities package karta hai. Woh skills ko outside apps ke connections ke saath bundle karta hai, aur kabhi commands aur sub-agents bhi shamil karta hai. Is tarah aik install hone wala package banta hai jo kisi job function ke liye hota hai.
Talluq aik aik line mein:
- skill tareeqa sikhati hai
- app ya connector outside system tak rasai deta hai
- plugin unhein aik package mein jor deta hai
Claude ab skills, connectors, aur plugins ko claude.ai/directory par aik hi directory mein dikhata hai. Apna banane se pehle dekhne ka yeh sab se tez tareeqa hai ke pehle se kya mojood hai.
Custom GPTs Ka Kya?
ChatGPT mein custom GPTs bhi hain.
Custom GPT alag configured assistant hota hai jise aap jaan boojh kar kholte hain. Us ki apni instructions, persona, knowledge, aur tools hote hain.
Yeh skill se mukhtalif cheez hai.
Mukhtasar muqabla:
- Project: kaam ke aik silsile ka context rakhta hai
- Skill: task karne ka tareeqa rakhti hai
- Custom GPT: aapko alag configured assistant deta hai
- Plugin: capabilities ko aik package mein jorta hai
Yaad rakhne wali line:
Projects knowledge rakhte hain. Skills tasks karti hain.
Customer-preparation skill project mein stored customer files istemaal kar sakti hai.
Project what deta hai. Skill how deti hai.
Aap mechanics ko Skills aur Connectors mein parhein ge.
Yaad rakhein: jab tareeqa baar baar aaye, use package karein.
Hissa 3: Rasai Barhana
8. Connectors, Apps, Aur Sawal Ko Sahi Rasta Dena
Ab tak assistant zyada tar us information ke saath kaam karta raha hai jo aap ne us ke saamne rakhi.
Lekin aapki zyada tar asli information kahin aur rehti hai:
- calendar
- cloud storage
- project-management tools
- chat systems
- company knowledge bases
Connectors aur apps assistant ko un systems tak pohnchne dete hain.
Connector Ya App?
Claude connector kehta hai.
ChatGPT usi product category ko app kehta hai. Pehle woh bhi connector kehta tha, phir naam badal diya.
Alfaaz mukhtalif hain. Maqsad aik hi hai.
Aapki di hui permissions ke mutabiq connector ya app assistant ko kisi aur system mein search, read, aur kabhi action lene de sakta hai.
Misaal ke taur par:
- "Woh email dhoondein jahan hum ne vendor contract par baat ki thi."
- "Kal meri kaun si meetings hain?"
- "Pichhle hafte ke project notes ka khulasa karein."
- "Mere sab se zyada priority wale tasks kaun se hain?"
Ab assistant aapke asli systems se jawab de sakta hai aur aapko har cheez chat mein copy nahin karni parti.
MCP: Mushtarik Integration Standard
Dono ecosystems Model Context Protocol, ya MCP, ko support karte hain.
Aam mukhtasar misaal AI tools ke liye USB-C hai: aik standard plug, taake aik baar bana tool kai mukhtalif assistants istemaal kar sakein. Claude ke custom connectors MCP par chalte hain, aur Apps SDK ke zariye ChatGPT ke apps bhi isi par bante hain.
Yeh waqai kaam ki baat hai, lekin yeh misaal aik cheez chhupa deti hai jo aapko nahin bhoolni chahiye. USB-C cable lagana trust ka faisla nahin hota. MCP tool connect karna hota hai. Plug standard hai; access standard nahin.
Connections Permission Ke Faisle Hain
Tool connect karna sirf productivity ka faisla nahin. Yeh security aur governance ka faisla bhi hai, yaani is baat ka faisla ke kis ko kya dekhne aur karne ki ijazat hai.
Kuch bhi connect karne se pehle poochein:
- Yeh kya parh sakta hai?
- Kya yeh outside world mein likh, bhej, delete, khareed, ya koi aur action le sakta hai?
- Kya mujhe source par bharosa hai?
- Kya mujhe yeh account ya system connect karne ki ijazat hai?
Work email connect karna bahut zyada information saamne la sakta hai, chahe assistant sirf wahi dekh sakta ho jo aapka apna account pehle se dekh sakta hai.
Connectors aur apps ko software install karne ki tarah samjhein: kaam ke hain, aur jaan boojh kar permissions ke gate ke peeche hain.
Search, Thinking, Ya Research?
Beginners aksar task ke liye ghalat mode chunte hain.
| Aapko kya chahiye | Kya istemaal karein | Misaal |
|---|---|---|
| Aik current fact | Web search | "Aaj ka exchange rate kya hai?" |
| Mushkil reasoning, koi nai information nahin | Thinking mode | "Pricing ki in teen strategies ka muqabla karein." |
| Mukammal, cited investigation | Research mode | "Warehouse robotics ki current market par research karein." |
| Aapki organization ki information | Workplace search ya connected apps | "Latest approved pricing policy dhoondein." |
Pehle yeh sawal poochein:
Mein assistant se kis qisam ka kaam maang raha hun?
Jis fact ka jawab search seconds mein de, us par deep research kharch na karein. Jab asal masla reasoning ho to search ki taraf na jayein. Jab kaam ko multi-source investigation chahiye ho to ordinary chat istemaal na karein.
Research Mode Mukhtalif Qisam Ki Cheez Hai
Research mode "behtar search" nahin hai.
Assistant investigation ka plan banata hai, kai searches chalata hai, suragh ke peeche jata hai, sources parhta hai, aur citations ke saath structured report deta hai.
Is mein seconds nahin, minutes lagte hain.
Aap quick sawal nahin pooch rahe, balke investigation delegate kar rahe hain.
Is ka yeh matlab bhi hai ke report aane par aapki judgment khatam nahin hoti. Aham claims check karein. Citations kholein. Research aapki rasai barhati hai. Aapki accountability kisi aur ko nahin deti.
Yaad rakhein: sawal bhejne se pehle use sahi rasta dein.
9. Use Us Kaam Par Sabit Karein Jise Aap Pehle Se Jaante Hain
Ab course ka sab se aham sawal:
Aapko kaise pata chale ke assistant aapke kaam mein waqai acha hai?
Benchmark par nahin.
Demo par nahin.
Aapke task par, aapke data ke saath, aapke standard ke mutabiq.
Beginner ke liye behtareen tareeqa seedha hai:
Assistant ko us kaam par test karein jo aap pehle hi mukammal kar chuke aur jis par bharosa karte hain.
Misaal
Aik program director har quarter attendance aur employment outcomes ka analysis karta hai.
Woh agle report mein AI ki madad chahta hai.
Use naye data par bharosa kar ke shuru nahin karna chahiye.
Is ke bajaye woh assistant ko pichhli quarter ka raw data deta hai, kyun ke use pehle se pata hai ke sahi analysis kaisa dikhta hai.
Phir woh assistant ko kaam dobara banane ko kehta hai.
Ab us ke paas muqable ke liye asli cheez hai.
Agar assistant aham pattern chhor de, to woh instructions behtar kar sakta hai.
Agar pata chale ke data mein analysis ke liye zaroori koi cheez hai hi nahin, to use data ka masla mil gaya.
Agar assistant phir bhi kaam ka koi hissa bharose se na kar sake, to woh hissa insaan ke paas rehta hai.
Teeno outcomes kaam ke hain.
Sabit Karne Ka Paanch-Step Loop
- Baar baar hone wala aik task chunein. Precise hon.
- Purani misaal dhoondein. Aisi jis ka sahi result aap pehle se jaante hon.
- Assistant se use dobara banwayein. Woh context dein jo aam taur par aapke paas hota.
- Muqabla karein. Kya match hua? Kya fail hua? Kaun si instruction missing thi?
- Behtar karein aur phir chalayein. Tab tak jaari rakhein jab tak kaam kaafi acha na ho, ya jab tak aap yeh natija na nikaal lein ke task delegate nahin karna chahiye.
Woh aakhri outcome failure nahin hai.
Yeh seekhna ke kya delegate nahin karna chahiye, aik ghante ke qabil hai.
Pass Hona Asal Mein Kya Sabit Karta Hai?
Yeh aapko milte-julte mustaqbil ke kaam ke liye kamaya hua aitmaad deta hai.
Yeh sabit nahin karta ke assistant mustaqbil ke har case mein durust hoga.
Aur yeh zimmedari transfer nahin karta. Aapko phir bhi:
- check karna hai ke naye results munasib lagte hain ya nahin
- final kaam ki zimmedari leni hai
- jahan munasib ho AI assistance disclose karni hai
- high-risk decisions ko durust insani control mein rakhna hai
Agar mumkin ho to wahi test Claude aur ChatGPT dono mein chalayein. Apni samajh wale task par do assistants ka muqabla platform awareness banane ka sab se tez tareeqa hai.
Yeh chhoti exercise aapki pehli evaluation suite hai.
Baad mein Checker Par Bharosa aur Eval-Driven Development isi aadat ko asal engineering discipline mein badalte hain.
Yaad rakhein: trust known work ke muqable mein kamaya jana chahiye, ache dikhne wale jawab se maan nahin lena chahiye.
Prompts Azmane Se Pehle Chhota Sa Khulasa
Ab aapke paas dono products ka naqsha hai.
Claude aur ChatGPT mukhtalif tools hain jo bahut se milte-julte ideas ke gird bane hain.
- Models aur thinking modes control karte hain ke aap kitni capability kharch karte hain.
- Attachments assistant ko woh context dete hain jo use isi waqt chahiye.
- Projects kaam ke aik silsile ka context aik jagah rakhte hain.
- Standing instructions stable rules rakhti hain.
- Memory aapke bare mein evolving context rakhti hai.
- Artifacts, writing blocks, aur code blocks mukammal kaam ko chat scroll se bahar rakhte hain.
- Skills baar baar istemaal hone wale methods package karti hain.
- Plugins aik poori qisam ke kaam ki capabilities bundle karte hain.
- Claude ke connectors aur ChatGPT ke apps external systems tak pohnchte hain.
- MCP dono ke neeche mushtarik integration standard hai.
- Search, thinking, aur research teen mukhtalif sawalon ke liye teen mukhtalif raste hain.
- Us kaam par testing jise aap pehle se jaante hain trust kamane ka tareeqa hai.
Products badalte rahein ge.
Yeh mental model current interface se bahut zyada arse tak chalna chahiye.
Abhi Azmayein: Chhe Prompts
Cockpit ke bare mein parhna us mein baithne jaisa nahin. Yeh chhe exercises taqreeban pachis minute leti hain.
1. Guided Tour Maangein
Chalane se pehle un do features ka andaza lagayein jin ka zikr assistant se tawaqqo hai.
Give me a guided tour of this workspace as it exists today.
Walk through:
- how I change models and when I should
- what happens when I upload a file
- what projects are for
- what you remember about me and where I control that
- where the research option lives
Keep it practical. One short paragraph per feature, and tell me which
features need a paid plan.
Kya note karein: ise apne andaze se, phir screen par asal mein nazar aane wali cheez se compare karein. Apne interface ke bare mein jahan woh ghalat ho, wahi sab se dilchasp hissa hai.
2. Aik Task Dono Assistants Mein Chalayein
Apne hafte ka koi chhota, asli task chunein: likhne ke liye paragraph, summarize karne ke liye table, ya chhota analysis.
Chalane se pehle andaza lagayein ke kaun sa assistant behtar kare ga, aur kyun.
Here is a real task from my work: [paste the task and its context].
Complete it.
Then, in three sentences, tell me what additional context would have
helped you do it better.
Kya note karein: outputs compare karein, phir compare karein ke har aik ne kya missing bataya. Doosra muqabla aam taur par pehle se zyada sikhata hai.
3. Apna Pehla Project Set Up Karein
Kaam ka aisa silsila chunein jahan aap wahi background pehle hi dohra chuke ya wahi documents dobara upload kar chuke hon.
I want to set up a project for this stream of work: [describe it].
Interview me briefly. Then give me two things I can paste into the
project settings:
1. project instructions covering my role, audience, tone, and standing rules
2. a priority-ordered list of five to ten documents I should upload as
project knowledge
Kya note karein: jo sawal woh aapse poochta hai, wahi context hai jo aap har hafte haath se dobara type karte rahe hain.
4. Yaad Rakhi Hui Cheezon Ka Audit Karein
Show me what you currently remember about me from our conversations,
as a plain list.
For each item I will tell you: keep, correct, or delete.
Then tell me where in settings I can manage memory myself.
Kya note karein: kya us list mein koi aisi cheez hai jo shared screen par aapko sharminda kare. Is exercise ki asli wajah yahi hai.
5. Teen Asli Sawalon Ko Rasta Dein
Apne asli kaam se teen sawal chunein:
- aik jis ke liye jaldi se current fact chahiye
- aik jis ke liye mushkil reasoning chahiye
- aik jo proper report ka haqdar hai
Har sawal se pehle rasta chunein: search, thinking, ya research.
Phir usi tareeqe se bhejein aur dekhein ke fit durust tha ya nahin.
Kya note karein: jo cheezein fit nahin huin. Aisi research run jo search honi chahiye thi, dono products mein sab se aam zaya hai.
6. Assistant Ko Apni History Par Sabit Karein
I want to test whether you can take over a recurring task of mine.
The task: [describe it precisely].
I am giving you a past example where I already know the correct result:
[attach the old input data or material].
Reproduce the analysis or output the way I would have done it.
I will compare your result with what I know to be correct, and we
will iterate.
I am holding back my original answer until we finish, so your work is
not shaped by it.
Muqabla mukammal hone par teen cheezein likhein:
- assistant ne kya match kiya
- use mazeed kis description ki zaroorat thi
- kya insaan ke paas rehna chahiye
Kya note karein: woh list aapka pehla evaluation record hai. Use sambhal kar rakhein.
Jaldi Se Apna Jaiza
Peeche dekhe baghair in ke jawab dein.
- Current model names yaad karne ke bajaye model tier pattern seekhna kyun behtar hai?
- Thinking mode kab istemaal karna chahiye?
- Instructions, memory, aur projects ko alag karne ka rule kya hai?
- Ordinary chat aur artifact ya writing block mein kya farq hai?
- Jumla mukammal karein: projects ___ rakhte hain; skills ___ karti hain.
- MCP kya hai, aur aik se zyada assistant istemaal karne par yeh kyun aham hai?
- Aapke pehle AI test mein aisa purana case kyun hona chahiye jis ka sahi result aap pehle se jaante hon?
- Us test ko pass karna aapko kya deta hai, aur kya kabhi transfer nahin karta?
Jawabat
- Model names aur versions tezi se badalte hain. Pattern (fast, thinking, flagship) zyada arse tak rehta hai aur aapko task ke mutabiq chunne deta hai.
- Multi-step reasoning, ehtiyat wale analysis, math, mushkil code, aur har us kaam ke liye jahan zyada reasoning intizaar ke qabil ho.
- Stable rules ke liye instructions, evolving context ke liye memory, scoped work ke liye projects.
- Ordinary chat conversation hai. Artifacts, writing blocks, aur code blocks alag work product rakhte hain jise aap edit aur alag se istemaal kar sakte hain.
- Projects knowledge rakhte hain; skills tasks karti hain.
- MCP Model Context Protocol hai, jo AI systems ko outside tools aur data se common interface ke zariye connect karne ka open standard hai. Yeh is liye aham hai ke integration ke concepts, aur aksar tools khud, aik product se doosre tak ja sakte hain.
- Kyun ke muqable ke liye bharosemand jawab chahiye. Known truth ke baghair aap nahin jaan sakte ke assistant ne kaam dobara banaya ya sirf aise dikhaya jaise banaya ho.
- Yeh milte-julte mustaqbil ke kaam ke liye tested confidence deta hai. Yeh accountability kabhi transfer nahin karta: aap phir bhi aham outputs verify karte aur final result ki zimmedari lete hain.
Agar sawal 8 par aap ruk gaye the, to Concept 9 ko AI Fluency ki Diligence competency ke saath dobara parhein. Dono mil kar is course ka professional core bante hain.
Is Course Mein Shamil Nai Istilahaat
Model picker. Model ya reasoning effort ka level chunne ke liye istemaal hone wala control.
Thinking mode, ya extended thinking. Aisa mode jo jawab se pehle model ko zyada reasoning effort deta hai. Dheema hai aur mushkil tasks par waqt ke qabil hai.
Project. Jaari kaam ke aik silsile ka workspace, jis mein chats, knowledge, aur instructions hoti hain.
Project knowledge. Project level par stored files, taake us project ki har conversation unhein istemaal kar sake.
Standing instructions. Stable rules jo conversations mein wasi taur par lagu hote hain: tone, role, output preferences.
Memory. Woh context jo assistant conversations ke darmiyan aapke bare mein rakhta aur baad mein istemaal kar sakta hai.
Incognito chat, ya temporary chat. Aisi conversation jo history aur memory se bahar rakhi jaati hai. Vendors safety ke liye limited waqt tak aik copy phir bhi rakhte hain, is liye yeh memory-free hoti hai, trace-free nahin.
Artifact. Claude mein alag output, jaise document, code file, diagram, page, ya interactive app, jo chat ke saath banta hai.
Writing block, code block. Current ChatGPT mein conversation ke andar editable region jo likhe hue ya coded kaam ko alag object ke taur par rakhta hai. Inhon ne canvas ki jagah li.
Skill. Instructions, examples, resources, aur kabhi code se bana packaged method, jo milta-julta task saamne aane par load hota hai.
Agent Skills standard. agentskills.io par open specification jo skill package karne ka portable tareeqa define karti hai aur jise kai mukhtalif vendors ke tools ab parhte hain.
Plugin. Aisa bundle jo poori qisam ke kaam ke liye skills, connectors, aur commands jaisi capabilities package karta hai.
Custom GPT. ChatGPT mein alag configured assistant jis ki apni instructions, persona, knowledge, aur optional tools hote hain.
Connector. Email, calendar, storage, ya kisi aur tool jaise outside system se permission ke saath connection ke liye Claude ka lafz.
App. External-tool integration ki usi category ke liye ChatGPT ka lafz. Pehle inhein connectors kaha jata tha.
Model Context Protocol (MCP). Open standard jo AI systems ko common interface se outside tools aur data ke saath connect karta hai.
Research mode. Multi-step mode jo investigation plan karta, sources mein search karta, aur cited report deta hai.
Enterprise Search. Claude ki organization-wide search jo connected company tools ko aik searchable knowledge base samajhti hai.
Yeh Aage Kahan Le Jata Hai
Ab aap Claude aur ChatGPT dono ka main chat workspace chala sakte hain.
Is se Foundations ka baqi hissa theory ke bajaye amli kaam ban jata hai.
Markdown In, HTML Out agla course hai. Yeh un zyada tar outputs ke peeche document layer samjhata hai jin se aap abhi mile hain.
Woh Code Jo Aap Kabhi Nahin Likhte samjhata hai ke jab aap ne software bayan kiya aur working code saamne aya to kya hua.
Skills aur Connectors Concepts 7 aur 8 mein shamil dono capabilities mein zyada gehra jata hai.
AI Ke Daur Mein Kaise Sochein us judgment ko mazboot karta hai jise Concept 9 ne amli proving method mein badla.
Baad mein Professionals Ke Liye Cowork aur OpenWork aur General Agents courses chat window se aage desktop aur agentic systems mein le jate hain.
Aur yahan istemaal kiya gaya saada proving loop Checker Par Bharosa mein sanjeeda engineering discipline ban jata hai.
Agar aap credentials jama kar rahe hain, to dono vendors muft product-learning resources publish karte hain, aur Certifications un exams ka naqsha deti hai jinhein yeh material support karta hai.
Sources Aur License Note
Yeh crash course is book ka original kaam hai. Product claims ko 25 August 2026 ko Anthropic aur OpenAI ki public documentation se verify kiya gaya tha.
Anthropic ka muft Claude 101 course aur OpenAI ke help resources ne product topics chunne mein rehnumai ki, lekin yeh course un ka text, structure, ya exercises dobara pesh nahin karta.
Features, names, plan limits, aur pricing aksar badalte hain. Jahan yeh page aur live product mukhtalif hon, vendor ki current documentation authority hai.
- Anthropic Help Center aur Claude documentation
- OpenAI Help Center
- Projects ke liye retrieval augmented generation (RAG), Anthropic Help Center
- Incognito chats istemaal karein, Anthropic Help Center
- Claude ki chat search aur memory istemaal karein, Anthropic Help Center
- Skills, connectors, aur plugins ko aik directory mein browse karein, Anthropic Help Center
- ChatGPT mein skills, OpenAI Help Center
- ChatGPT mein apps, OpenAI Help Center
- ChatGPT mein projects istemaal karna, OpenAI Help Center
- Skills, connectors, aur plugins ki Claude directory
- Model Context Protocol
- Agent Skills open standard
Flashcards Se Parhai Mein Madad
Apni Samajh Test Karein
Yeh scenarios aapko kisi aur ke workspace mein le jate hain jahan faisla pehle se aapka intizaar kar raha hai. Product name ke bajaye pattern se jawab dein, aur note karein ke har scenario asal mein kaun sa concept test kar raha hai.