[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]

[@initlayers](/creator/twitter/initlayers)
"24 October 2025 Revised LLM basics stages of building LLMs OOPs - mostly around classes and objects Project: Finally resumed. I built core modules to test project feasibility. I then downloaded Ollama and tested TinyLlama locally. Obviously it's not very great but with my GPU and RAM I am satisfied with anything I can possibly get. The next couple of days will be intense. I will try to code stuff from scratch"  
[X Link](https://x.com/initlayers/status/1981835386201727448) [@initlayers](/creator/x/initlayers) 2025-10-24T21:29Z XXX followers, XXX engagements


"Ever wondered what people mean when they say "context window" in LLMs Think of an LLM like a human with limited short-term memory. The context window is that memory span - it decides how much text the model can "remember" while generating a response. GPT-3 had a context windows of 2K-4K tokens. GPT-4 Turbo 128K. Claude XXX Even more. Larger context = more room to think recall and reason. You can feed entire books conversations or codebases and still get coherent answers. But here's the catch - a bigger window doesn't mean perfect memory. LLMs don't store information like humans they re-read"  
[X Link](https://x.com/initlayers/status/1983390838764851374) [@initlayers](/creator/x/initlayers) 2025-10-29T04:30Z XXX followers, XXX engagements


"Your text doesn't go straight into the model. It first passes through a tokenizer - a system that breaks words into small chunks called tokens. Example: understanding - under + standing LLMs don't read words. They read numbers"  
[X Link](https://x.com/initlayers/status/1983775019294667249) [@initlayers](/creator/x/initlayers) 2025-10-30T05:56Z XXX followers, XX engagements


"Each token depends on everything before it. That's why longer prompts slow it down - it's carrying your whole chat history each time. This running memory is called the context window. GPT-4 Turbo can handle up to 128000 tokens (around a 300-page book)"  
[X Link](https://x.com/initlayers/status/1983775053100839169) [@initlayers](/creator/x/initlayers) 2025-10-30T05:56Z XXX followers, XXX engagements


"@krishdotdev Selective consumerism and extremely high product consciousness. India has all of it. Accounting for each rupee spent"  
[X Link](https://x.com/initlayers/status/1983113880084922534) [@initlayers](/creator/x/initlayers) 2025-10-28T10:09Z XXX followers, XXX engagements


"@OjasSharma276 I hope you are okay. What did you apply for though Like role"  
[X Link](https://x.com/initlayers/status/1983444577823629336) [@initlayers](/creator/x/initlayers) 2025-10-29T08:03Z XXX followers, XX engagements


"Once tokenized your request turns into a long list of numbers. This is sent to OpenAI's servers (or whichever model host you're using). Every message carries context - everything you typed so far. That's how ChatGPT "remembers" your conversation"  
[X Link](https://x.com/initlayers/status/1983775036847861808) [@initlayers](/creator/x/initlayers) 2025-10-30T05:56Z XXX followers, XX engagements


"Now the real magic begins. Inside giant clusters of GPUs the model looks at your tokens and predicts what comes next. one token at a time. The entire essay you got back Generated word by word. Prediction by prediction"  
[X Link](https://x.com/initlayers/status/1983775043743313952) [@initlayers](/creator/x/initlayers) 2025-10-30T05:56Z XXX followers, XX engagements


"As each token is predicted it's streamed back to you in real time. That's why you see words appear one by one. You're literally watching the model "think.""  
[X Link](https://x.com/initlayers/status/1983775072973410695) [@initlayers](/creator/x/initlayers) 2025-10-30T05:56Z XXX followers, XX engagements


"When the reply ends your entire chat (your message + its answer) is stored temporarily for continuity. Once the context window fills up older parts are forgotten. ChatGPT doesn't have memory. It has context"  
[X Link](https://x.com/initlayers/status/1983775082515460495) [@initlayers](/creator/x/initlayers) 2025-10-30T05:56Z XXX followers, XX engagements