First Steps - Nail the fundamentals!

 Every new learning feels like a mountain to start with, but the beauty of learning is you can break it down and start with fundamentals. Once you learn the fundamentals, you gain confidence, and slowly what used to feel like a big mountain is now achievable, but again it will lead to another mountain, which is the interesting part.

Similarly, with AI, there is so much discussion, so much material, and so many tools also to learn. I want to start this journey by just asking questions and then going and finding the answers to those questions. I will take help of AI, but then I will summarize my understanding back in this blog.

To start with, some of the questions that come to my mind are as follows:

• What is Gen AI?
• Why is it so special?
• How is it different from other forms of AI?
• What is LLM?
• Why are different models trained on different models?
• Volume of data. What stops or removes that?
• How many more models will we see, and do we see a pattern in the release of features in each of those models?
• What is an example of a feature in a model?
• What are the benchmarks?
• How do we measure the power of a model or the success of a model?
• What happens when we ask or send a prompt, for example in ChatGPT or any AI application?
• The roles of RAG, knowledge base, knowledge graph in Gen AI.
• What is embedding?
• What is positional embedding?
• How does it affect the output?
• What are all the factors that affect the output of your prompt? One of the factors could be the way you prompt also.
• Why do models hallucinate?
• Can we control models, or should we control the application built on models or the agents?
• What is unique about an AI agent?
• What is an agentic AI platform?
• What characteristics define an agentic AI platform?
• What are the limitations of agentic AI platforms?
• How to build an AI agent? 
• What characteristics are there for an AI agent? How is it different from a chatbot? 
• Role of guardrails. Are there different types of guardrails? 
• What is eval? What is Ragas framework? 
• How much can we trust the output of an agent? Are there known patterns of failures? 

A New Blog - A New Identity

After close to 15 years of blogging at www.enjoytesting.blogspot.com, I switched to more posts on LinkedIn and stopped blogging. 

Now, I feel that I want to blog again but on a totally different theme - AI in Testing. 

There are more than 20 tools that get discussed on a weekly basis.
There are so many courses and roadmaps for learning about AI. 
There are so many questions as well in everyone's mind.

Every day, there is so much new that is coming in the AI world that it is super difficult to catch up to the news. But we cannot be left behind. This blog is my attempt to consolidate my learning journey, experiences, questions, insights and everything related to AI.

Follow along, comment your thoughts and I leave everything to The Compound Effect. 

Some of the raw thoughts in my mind are as follows:

- Learn
- Practice
- Implement

• Fundamentals of AI and GenAI
• Concepts of AI: Token, Agent, Agentic Workflow, Hallucination, Context Condensing, Prompt Caching, RAG
• A bit of history - how things have evolved
• A bit of fundamentals - Machine Learning, Deep Learning, Neural Networks
• Aishwarya Courses (Text)
• Claude, Copilot, Codex, CLi Courses/Documentation
• MCPs, CLI, Knowledge Graph
• Different Tools - Wispr Flow, Copilot (Free vs Enterprise), Claude Code, Kiro, Cursor, 
• Building Agents
• AI Evals, RAGAs
• Application - Using Playwright, Appium MCP
• AI in QE
• Winning with AI, Human Edge in AI, Governance, Explainability, Anthropic - Latest Topics
• Models - LLMArena (Leaderboard), How they evaluate or score models
• How to Test AI - AI Chatbots - How to bring determinism inside non-deterministic space
• How to automate AI Chatbots
• Guardrails
• Few Free Videos on LLMs, Fundamentals
• Your point of view on how to use AI

What are your thoughts on the above?