AI Explained · Episode 2
ChatGPT Bhool Kyun Jaata Hai? Context Window Explained in Hindi
Why a long ChatGPT chat forgets your first messages: the context window is AI's desk. Plus 3 tricks to stop it forgetting.
- AI Fundamentals
- Beginner
- 2 min read
- Video posted 2 Nov 2026
- Completed
Short mein: ChatGPT bhoolta nahi, uski desk bhar jaati hai. Ise context window kehte hain. Desk saaf rakho, aur zaroori baat shuru ya aakhir mein likho.
Context window = AI ki working desk
AI ke paas ek desk hai. Jo is desk par rakha hai, AI bas utna hi dekh sakta hai. Is desk ka naam hai context window.
Sabse bada raaz: AI ko kuch yaad nahi rehta
Model ke paas ek message se doosre message tak koi memory nahi hoti. Har baar jab tum send dabate ho, app poori chat shuru se dobara desk par rakhti hai, aur AI use phir se padhta hai.
(ChatGPT ka “Memory” feature kuch notes save karta hai, par wo notes bhi isi desk par hi rakhe jaate hain.)
Desk bhar gayi to?
Lambi chat mein desk bhar jaati hai. Tab sabse purane messages AI ko dikhne band ho jaate hain: kuch apps unhe hata deti hain, kuch unka summary bana deti hain. Yahi hai AI ka “bhoolna”. Ye bug nahi, desk ka size hai.
Desk par kya-kya rakha hota hai
- Hidden instructions (system prompt), jo app khud jodti hai
- Tumhare saare messages
- Files jo tumne daali
- Aur AI ke apne jawab bhi
Desk ka size tokens mein naapa jaata hai, wahi bricks jo Episode 1 mein dekhe. Example: GPT-4o ka context window 1,28,000 tokens ka hai. Ek token lagbhag 0.75 English words, to ye karib 96,000 words, yaani lagbhag 300 pages ki kitaab. Kuch models 10 lakh tokens tak lete hain.
Aur yaad hai? Hindi mein tokens zyada bante hain, isliye Hindi chat desk jaldi bharti hai.
Desk badi kyun nahi kar dete?
Standard “self-attention” mein har token baaki har token se juda hota hai. Tokens doguna, to jode lagbhag chaar guna, aur kaam bhi lagbhag chaar guna. Naye models isse tez tareeke bhi use karte hain, par badi desk phir bhi mehngi padti hai.
Surprise: Lost in the Middle
Badi desk par bhi AI beech ki baatein zyada miss karta hai. Research ne dikhaya ki lambe input ki shuruaat aur aakhir ki baatein AI behtar use karta hai, beech ki kam. Naye models isme behtar hue hain, par advice kaam ki hai.
3 tricks
- Sabse zaroori instruction prompt ki shuruaat ya aakhir mein likho, beech mein nahi.
- Naya topic, nayi chat. Ek hi chat mein sab kuch mat chalao.
- Lambi chat ka summary banwao, aur use nayi chat mein paste karo.
Ek line yaad rakho: AI bhoolta nahi, uski desk bharti hai. Desk saaf rakho.
Try it: related prompts
-
Role, Context, Task, Format: the 4-Part Prompt Template
The fill-in template behind every prompt on the PROMPT sheet. Tell AI its role, your situation, the exact task and the shape of the answer.
- Prompt Engineering
- ChatGPT
- Any chat AI
- Beginner
Tum [ROLE] ho. [CONTEXT: main kaun hoon, situation kya hai]. [TASK: exactly kya chahiye]. [FORMAT: bullets, table, ya kitne words].
-
Answer Only From My Document
A copy-paste line from the SACH sheet. Paste your text, add this line, and AI answers only from it, or says it is not in the document.
- Fact-checking
- ChatGPT
- Any chat AI
- Intermediate
Sirf neeche diye text se jawab do. Text mein nahi hai to bolo: 'document mein nahi hai'.
Quick check: 3 questions
Samjha ya sirf padha? Answers sirf tumhe dikhte hain.