<![CDATA[Short intro to GPT]]>https://devesh001.hashnode.devRSS for NodeSun, 08 Nov 2026 09:12:37 GMT60<![CDATA[Agentic AI Made Simple: What Are Agents and How Do They Work?]]>https://devesh001.hashnode.dev/agentic-ai-made-simple-what-are-agents-and-how-do-they-workhttps://devesh001.hashnode.dev/agentic-ai-made-simple-what-are-agents-and-how-do-they-workMon, 18 Aug 2025 12:59:32 GMTAgentic AI is a type of artificial intelligence that acts like a helper or worker for you. Let’s talk about what agents are and how they work, using very simple words.

What Is an "Agent"?

An agent is like a robot, computer program, or app that:

  • Sees or hears what’s happening around it

  • Thinks about what to do

  • Does something

You can think of your phone’s voice assistant—like Siri or Alexa—as an agent. It listens to you, figures out what you want, and then helps you do it.

How Does an Agent Work?

Here’s how easy it is:

  1. Looks or listens: The agent uses sensors or cameras to see, or microphones to hear, what’s around.

  2. Thinks: It makes decisions based on what it finds out.

  3. Acts: The agent does something—like speaking, turning on lights, or answering questions.

Simple Examples

  • Smart lights: Lights that turn on when you walk into a room. The light is the agent. It "sees" you, "thinks" you want light, and then "acts" by turning on.

  • Chatbots: Chatbots are computer programs that answer your questions online. They "read" your words, "think" about what you mean, and "act" by replying.

  • Robot vacuums: These devices clean your house. They "see" dirt, "think" about how to clean it, and "act" by moving around and vacuuming.

Why Is Agentic AI Useful?

Agentic AI helps make life easier. It can do things for you without you having to ask every single time. These smart agents help with cleaning, answering questions, and many more things.

]]><![CDATA[What is Chain-of-Thought (CoT) Reasoning in GPT?]]>https://devesh001.hashnode.dev/what-is-chain-of-thought-cot-reasoning-in-gpthttps://devesh001.hashnode.dev/what-is-chain-of-thought-cot-reasoning-in-gptFri, 15 Aug 2025 09:53:56 GMTChain-of-Thought reasoning refers to a process where a model doesn't just output an answer, but instead breaks down its reasoning into a series of logical steps. It mimics how humans solve problems: step by step, reflecting, and refining until reaching a solution.

Example:
Instead of just answering “24” to “What is 6 × 4?”, a thinking model with CoT might say:

  • First, I know that multiplication is repeated addition.

  • 6 × 4 means adding 6 four times: 6 + 6 + 6 + 6.

  • Adding up, that’s 24.

Why is CoT Useful for Thinking Models?

  • Transparency: You can see how the answer was reached.

  • Improved Accuracy: Breaking problems down can reduce mistakes.

  • Learning & Teaching: CoT outputs help users understand concepts.

Applications of CoT Thinking Models

  • Education: Teach students problem-solving.

  • Coding Help: Stepwise debugging and code explanations.

  • Business Analysis: Transparent decision-making processes.

]]><![CDATA[Vector Embeddings: Turning Data into Meaningful Numbers]]>https://devesh001.hashnode.dev/vector-embeddings-turning-data-into-meaningful-numbershttps://devesh001.hashnode.dev/vector-embeddings-turning-data-into-meaningful-numbersTue, 12 Aug 2025 12:04:01 GMTVector embeddings are numerical representations of complex data—such as words, sentences, images, or users—in a continuous, high-dimensional space.By converting unstructured inputs into arrays of numbers (vectors), embeddings enable machine learning models to process, compare, and manipulate diverse data using mathematical operations

Why do we need this?

Because computers don’t understand words like “cat” or “love” directly — they only understand numbers.

So we turn them into numbers in a smart way, so that:

  • Similar things are close together on the map
    Example: “cat” and “dog” will be near each other.

Vector embeddings are just number-based coordinates for words, sentences, or images — so that computers can understand meaning and find similarities, just like finding nearby places on a map.

]]><![CDATA[Tokenization in GPT: Breaking Text into Digestible Pieces]]>https://devesh001.hashnode.dev/tokenization-in-gpt-breaking-text-into-digestible-pieceshttps://devesh001.hashnode.dev/tokenization-in-gpt-breaking-text-into-digestible-piecesTue, 12 Aug 2025 11:21:23 GMTTokenization in GPT is the crucial process of converting human text into smaller units called tokens, which serve as the fundamental building blocks that language models can understand and process.

Tokens are the smallest units of text that GPT models process.

  • Individual words like "hello" or "computer"

  • Single characters like "a" or "?"

  • Punctuation marks like "." or ","

  • Spaces between words

For example, the sentence "I love programming!" might be tokenized into: ["I", " love", " programming", "!"]

]]><![CDATA[Stop Saying 'AI' - Here's What GPT Actually Means]]>https://devesh001.hashnode.dev/stop-saying-ai-heres-what-gpt-actually-meanshttps://devesh001.hashnode.dev/stop-saying-ai-heres-what-gpt-actually-meansTue, 12 Aug 2025 04:59:21 GMTGPT or Generative Pre-trained Transformer is a advanced Neural Network Models designed for understanding and generating human like text. This was developed by OpenAI.

Let’s break down the term GPT :

  • Generative: the models can create new content, generate text like human.

  • Pre-trained: these models are heavily trained by the engineers for doing specific task.

  • Transformer: they were introduced by Google in 2017 for making Google Translate, think transformer like a smart machine that can understand text the same way humans do, but much faster. It also learns context by tacking the relationship in data.

Resource - Google Published Doc

Real world applications powered by transformers:
- Chatbots, Language translation, writing assistants

Developer? Let’s Connect

]]>