Tech Tips, Tricks & Trivia

Vibe coding is a term popularized by Andrej Karpathy and the developer community to describe a shift where the "coder" focuses on high-level intent, design, and iterative prompting (the "vibe"), while the AI handles the syntax, boilerplate, and implementation. Vibe coding is a popular choice for prototyping, building MVPs, and driving fast-paced innovation, but the industry still has reservations about its code quality and security. That's not stopping AI Provider companies from targeting their products at vibe coders with their official documentation providing helpful guidance. In 2025 and 2026, "Vibe Coding" moved from a Twitter/X joke to official documentation. * Google: Now has a dedicated " Vibe Code with Gemini " landing page in AI Studio, explicitly using the term in their marketing. Google Cloud offers several tools for vibe coding .  * Replit: Officially calls itself the "#1 Vibe Coding Platform" and features engaging artic...

Grok can do the Hindi-English mix very well.  Grok's explanation and examples for the word "condescending" -  Arre yaar, "condescending" ka matlab hai – woh attitude jab koi upar se neeche dekhta hai, jaise woh khud Mount Everest pe baitha ho aur tum road ke pothole mein. Simple English: Talking to someone like they're a 5-year-old who just asked why the sky is blue, even if they're a grown adult with a PhD. Desi example for crystal clarity: Imagine your rich NRI cousin comes home after 10 years, pats you on the head and says, "Aww, still using an Android phone? So cute yaar, here take my old iPhone 12." That smug, "I know better than you" tone? Peak condescending behavior. Universal example: When someone explains something you already know, but sloooowly and loudly, like you're hard of hearing. "Do... you... know... what... a... smartphone... is?" Bonus Indian uncle version: Boss at office: "Beta, ye Excel ka formu...

Nikita Namjoshi 's video " How do thinking and reasoning models work? " from the Google for Developers channel explains the concepts behind "thinking models" or "reasoning models," such as Gemini, and how they use more computation at inference time to achieve better results in complex tasks. Video Summary The video focuses on how Large Language Models (LLMs) can be improved to handle complex tasks like coding, advanced mathematics, and data analysis by utilizing more compute power during the generation phase (inference or test time). The Problem: LLMs generate responses by predicting one token at a time. When solving complex problems, the model has to figure out the entire solution in a single pass to generate the correct final answer, which is difficult. The Solution: Chain of Thought (CoT): CoT prompting is a technique where the model is prompted to generate a series of intermediate steps (a "chain of thought") that lead to the final answer...

Today's AI coding tools look like fierce competitors, but they're all built on each other's work.  Today’s AI-powered IDEs like Cursor, Windsurf, Kiro and many others all stand on the giant shoulders of open-source work contributed by companies that once battled each other in the browser wars.  Windsurf, for example, is a fork of Microsoft’s VS Code. VS Code itself runs on Electron, which is built on top of Chromium and Node.js. Chromium comes with Google’s V8 JavaScript engine and uses Blink, a fork of Apple’s WebKit.  So even when Google, Microsoft, GitHub, and Apple compete publicly, their code ends up powering each other’s tools behind the scenes. Open source has turned “tech rivals” into “reluctant collaborators,” whether they planned it or not. The AI Model Dependencies The same pattern exists with AI models. Competing IDEs use each other's models, hosted on each other's clouds. GitHub Copilot (Microsoft) uses : OpenAI models (GPT-4.1, GPT-5 family) - hosted o...

21 Lessons 00:00:00 - Introduction to Generative AI and LLMs [Pt 1] 00:10:36 - Exploring and comparing different LLMs [Pt 2] 00:31:34 - Using Generative AI Responsibly [Pt 3] 00:40:54 - Understanding Prompt Engineering Fundamentals [Pt 4] 01:04:08 - Creating Advanced Prompts [Pt 5] 01:21:06 - Building Text Generation Applications [Pt 6] 01:36:37 - Building Chat Applications [Pt 7] 01:50:10 - Building Search Apps Vector Databases [Pt 8] 02:08:08 - Building Image Generation Applications [Pt 9] 02:30:32 - Building Low Code AI Applications [Pt 10] 02:46:55 - Integrating External Applications with Function Calling [Pt 11] 02:56:06 - Designing UX for AI Applications [Pt 12] 03:08:38 - Securing Your Generative AI Applications [Pt 13] 03:16:49 - The Generative AI Application Lifecycle [Pt 14] 03:29:12 - Retrieval Augmented Generation (RAG) and Vector Databases [Pt 15] 03:40:03: - Open Source Models and Hugging Face [Pt 16] 03:51:32 - AI Agents [Pt 17] 03:59:26 - Fine-Tuning LLMs [Pt 18] Relate...

Kaggle and Google are running a 5-day Gen AI Agents Intensive course with daily assignments and lectures from November 10 to 14, 2025.  Start with the  self-paced learning guide  or just check the highlights & key links below. Also see -  Kaggle's 5-Day Gen AI Intensive Course - Notes Notes - "It's not the year of the agent. It's the decade of the agent." - Andrej Karpathy Introduction to Agents Building successful agents isn't just about having the smartest model. The Agent is the combination of the model for reasoning, the tools for action and the orchestration layer managing that loop. Success hinges on the architecture, governance, security, testing, observability An agent can do more than just respond to a prompt — it can take actions to find information or get things done. Whitepaper:  Introduction to Agents Codelabs: Build your first agent using Gemini and ADK - From Prompt to Action Build your first multi-agent systems using ADK - Agent Architec...

Dheemanth Reddy and Bharath Kumar from Maya Research AI have developed an open-source, world-class voice conversational models  from their SPC Bengaluru office in HSR Layout. Their 3B parameter Maya1 model is open-weight and was trained entirely using free compute credits. Anyone can download the model, inspect it, use it, or fine-tune it. It’s not locked or proprietary like many large AI models. The model can understand speech and respond in natural sounding speech — similar to Siri, Alexa, or GPT voice chat. The model is strong enough to be compared with leading voice AI models, despite being built with minimal resources. Core Capabilities: 20+ emotional styles (e.g., cheerful, calm, dramatic). Zero-shot voice design - clone or design new voices without needing training data. 3B-parameter architecture, optimized for production-ready real-time streaming. Apache 2.0 license - businesses can deploy and monetize with no per-usage fees . Supports fine-tuning to create un...

The Google Generative AI Leader Certification Exam tests your skills in applying generative AI for business transformation and innovation. To prepare for the exam, review the study guide  [PDF] and follow the Learning Path on Cloud Skills Boost Update 26-Jan-2026: I passed the exam! I found the Generative AI Leader Learning Path on Cloud Skills Boost with its 5 modules and 100+ quiz questions and answers with reasoning within it both helpful and interesting. The exam is of 90 minutes and includes 45-60 multiple-choice questions. The certification is valid for 3 years. Check this playlist  Related Google Services & Resources - Google Search: Find information faster and with more accuracy Google Workspace with Gemini: Draft emails in Gmail, generate presentations in Slides, summarize meeting notes in Docs, and automate tasks in Sheets. Gemini App: Access Gemini directly through a dedicated app.  Gemini for Google Cloud: Build applications and services on Google Cloud...

Google recently published a report on the The ROI of AI 2025 - How agents are unlocking the next wave of AI-driven business value. AI agents are specialized LLMs that have specific roles, context, and objectives to independently plan, reason, and perform tasks with access to data function call APIs and can interact with other AI agents if needed. These can be pre-built or in-house built agents There is a fundamental shift from AI that simply assists to AI that can operate independently under human control and guidance.  We’re in the agentic era, where AI agents can independently execute tasks and make decisions— under human guidance and guardrails.  Levels of AI agent maturity Level 1 Simple tasks Chatbots Information retrieval Image generation Level 2 AI agent applications Customer service AI agents Creative agents Level 3 Multi-agent workflows Agentic workflows Agent orchestration Top gen AI impacts across business areas - Productivity Customer experience Business growth M...

This Week I Learned -  *  Confluent Cloud is a fully managed, cloud-native Kafka service that provides features like on-demand cluster creation, data generation, real-time stream processing, and account management. It is deployable within seconds, accessible across all major public clouds, and offers a free start with cost-effective pay-as-you-go pricing that scales with your growth. * Azure Local, formerly Azure Stack HCI, is a comprehensive, Azure-integrated hyperconverged infrastructure (HCI) solution that leverages Hyper-V as a component within a broader, cloud-centric framework.  * Data centers and cloud computing are responsible for 1 percent of the world’s energy-related greenhouse gas emissions * OpenAI has become one of the world's biggest renters of cloud servers * Roo Code is an AI-powered autonomous coding agent that lives in your editor. It gives you a whole dev team of AI agents like coding partner, a system architect, or specialized roles like a QA e...

This Week I Learned -  *  Google Cloud Generative AI Leader certification exam is meant for anyone in any job role, with or without hands-on technical experience. The exam is MCQ type and has a validity period of 3 years. * Comet is an AI-native browser launched by Perplexity in July this year. It combines traditional web browsing with real-time AI assistance that can read, summarise and act on live data. Currently, Comet is available to Perplexity Max customers on an invite-only basis. * Perplexity offered to buy Chrome, Google’s web browser, for $34.5 billion. * Humanize AI can transform your AI-generated content into natural, human-like text. This AI-to-human text converter converts output from ChatGPT, Bard, Jasper, Grammarly, GPT4, and other AI text generators into text indistinguishable from human writing. * Like a human brain, artificial neural networks take some kind of input (words, say) and produce outputs (other words). For the human brain, what happens in the...

 A list of AI coding assistants some of which come along with online editors - 1. GitHub Copilot   2. Cursor 3. Windsurf  4. Claude Code 5. Replit   6. v0 from Vercel 7.  Bolt 8. Base44 9. Firebase Studio 10. Lovable  - suited for non-coders 11. Kiro  (preview) is an agentic IDE  built and operated by a small, opinionated team within AWS . Kiro turns your prompt into structured requirements, design, and tasks that are then implemented by agents. Kiro’s agent hooks help you scale your work by delegating tasks to agents that run in the background, such as updating docs, generating unit tests, or optimizing your code for performance. Also see - A comparison of AI coding assistants

This Week I Learned -  * The Web Platform Status web site allows web developers to query and track features shipped in major browsers. *  Appwrite.io  is an open-source, all-in-one development platform that provides built-in backend infrastructure and web hosting, all conveniently managed in one place.  *  GitHub Apps are tools that extend GitHub's functionality. GitHub Apps can do things on GitHub like open issues, comment on pull requests, and manage projects. They can also do things outside of GitHub based on events that happen on GitHub. For example, a GitHub App can post on Slack when an issue is opened on GitHub. Once you authorize a GitHub App, the app can act on your behalf. The situations in which a GitHub App acts on your behalf vary according to the purpose of the GitHub App and the context in which it is being used. For example, an integrated development environment (IDE) may use a GitHub App to interact on your behalf in order to push changes you ...

Whats In A Name is a simple Flask app that uses the Sonar API from Perplexity.ai to fetch interesting, cultural, and historical facts about names. Enter a name, get a fun fact. That’s it! If you're getting started with Perplexity’s Sonar API, this is a very basic example you can try by just placing your own Perplexity Sonar API key at the point specified in the code shared on GitHub . Based on the scenario, you can use any of the following Sonar models: sonar (lightweight search)  sonar-pro (advanced search) sonar-deep-research (exhaustive research) sonar-reasoning (fast reasoning) sonar-reasoning-pro (premier reasoning) You can deploy Flask apps on the web using one of the many options like Anaconda's  PythonAnywhere . With a limited Beginner account, you can host one web app at your-username.pythonanywhere.com with restricted outbound Internet access from your apps at low CPU/bandwidth. 

This Week I Learned -  * After OpenAI’s $3 billion Windsurf deal collapsed, Google invested $2.4 billion while also hiring its CEO Varun Mohan and key staff. Cognition, the AI startup behind the viral AI coding agent Devin, has officially acquired Windsurf in a deal that will include the coding company's IP and user base. *  AI Agent School - An interactive educational platform for understanding AI agents * S Anand created an Amar Chitra Katha-style comic story based on his Vipassana experience  and also shared how he prompted ChatGPT to generate it. *  Agents (LLMs running tools in a loop) can think only for ~6 min * Nvidia controls more than 80% of the market for the chips used for building AI systems *  Adrien Friggeri  has been running every day for the past 15 years & making beautiful charts from the data he collects. * Micronations, sometimes also referred to as model countries and new country projects, are small, self-proclaimed entities ...

This hour-long video covers not just the maths behind LLMs but also the history. Complex topics are well explained.  * Training a large language model is like constructing a skyscraper, where every   brick is placed by an army of specialists working around the clock. But here, the bricks are data, the mortar is mathematics, and the blueprint is a complex interplay of algorithms.  *   LLMs don't truly understand. They excel not through cognition, but through colossal computation. *  Mathematics is the invisible backbone of these 'thinking machines'. *  N-grams - A statistical method used by early language models that broke down text into small sequences of words to predict the next word based on common combinations in a dataset. *  Transformers - A groundbreaking architecture introduced in 2017 that revolutionized the field by enabling machines to grasp context at an unprecedented scale , allowing models to pay attention to all parts of a se...

The LinkedIn Learning short course Building AI-Ready Applications with Azure Databases and AI by Muazma Zahid, Principal Group PM Manager at Microsoft provides an overview of creating AI-ready applications using Azure's data and AI services. The code related to the course is in GitHub Codespaces .  Highlights: Key services Azure offers - Azure AI Search is a search as a service solution that integrates AI capabilities to enhance search results. It can analyze and understand the content of your data, providing more relevant and accurate search results.  Azure AI Studio is an integrated development environment for building, training, and deploying AI models.  Azure Data Services provide the infrastructure for storing, processing, and analyzing large amount of data. Azure offers a variety of database services, including Azure SQL Database , Cosmos DB , PostgreSQL , and MySQL . These databases are optimized for AI workloads providing high performance scalability and sec...

Race to Certification 2025 learning and certification challenge running from July 1 to October 31, 2025 offers an opportunity to earn free professional certification exam attempts for the following exam titles: AI   Oracle Cloud Infrastructure 2025 AI Foundations Associate | 1Z0-1122-25 Oracle Cloud Infrastructure 2025 Data Science Professional | 1Z0-1110-25 Oracle AI Vector Search Professional | 1Z0-184-25 Oracle Cloud Infrastructure 2025 Generative AI Professional | 1Z0-1127-25 OCI  Oracle Cloud Infrastructure 2025 Foundations Associate | 1Z0-1085-25 Orac...

This Week I Learned -  * Gemini CLI is an open-source AI agent that brings Gemini directly into your terminal, with MCP support for extensibility and Human in the Loop for oversight.  * Prompting only changes how the model responds. Fine-tuning changes the weights—but it’s costly, brittle, and still static. Neither gives the model new knowledge in real-time. With Retrieval-Augmented Generation (RAG) , you “Retrieve” the most relevant data for the user’s query from your database, use this data to “Augment” the prompt you send to the LLM, and then let it “Generate” a response based on the user query + prompt + retrieved data. RAG doesn’t modify the model; it modifies the input. It retrieves relevant context (docs, tickets, policies) and feeds it into the prompt at inference time. No retraining. Just better answers, grounded in your own data. *  Google Apps Script Web Apps can act as powerful servers for open protocols like Model Context Protocol (MCP) and Agent2Agent (A2A)...

See more of my AI co-creations The idea for this cartoon was inspired by a line from a New York Times article : "...perseverance is a defect. If you have to struggle, it means you’re not good enough."