Engineering matter in AI

 

As much as I love getting my hands dirty with the latest AI frameworks, I’m also a big believer in intentional design. Here’s the mindset I bring to the table beyond just the code.

We’ve all seen it. You type a prompt, the AI spits out 200 lines of code, and—magically—it runs. It feels like a superpower. This is "Vibe Coding." It’s fast, it’s intoxicating, and it’s a brilliant way to prototype an idea in a weekend.

But as the above image from social media illustrates, there is a massive difference between a structure that looks like a house and one that can actually weather a storm.

---

The Allure (and Danger) of the Vibe

In the left panel, we see the "AI Vibe" house. On the surface, it has a roof and walls. But look closer:

• The Foundation: It’s sitting on a pile of loose rocks. In software, this is the lack of a scalable architecture or a proper database schema.

• The Logic: There’s a toilet in the middle of the kitchen. This is what happens when AI generates features without understanding the user flow or business logic.

• The Infrastructure: Plumbing is hanging out in the open, tangled and messy. This is your "spaghetti code"—untested, undocumented, and impossible to maintain.

Vibe coding is great for exploring, but it's dangerous for building. That's why Engineering matters in the age of AI!

---

The Engineer-Guided Evolution

The right panel shows the future of our industry: Engineer-Guided AI. Being an engineer in 2026 isn't about writing every semicolon; it’s about being the Architect of the Intelligence. When you guide the AI with an engineering mindset, you provide:

1. Structural Integrity: You define the patterns (MVC, Microservices, Clean Architecture) before the first line is generated.

2. Safety & Security: You don't just ask for a "login page"; you enforce OAuth standards, input validation, and encryption.

3. The Basement (Scalability): You ensure the foundation—the infrastructure and DevOps pipeline—is poured deep enough to support future growth.

---

The Bottom Line

AI is the most powerful power tool ever invented, but a power tool in the hands of someone without a blueprint just creates a faster mess.

Don't just code for the vibe. Build for the future. The world doesn't need more "houses" that collapse under pressure; it needs engineers who use AI to build cathedrals.

In AI world, follow your engineering passion to add business/human values

Moltbot

What is Moltbot / OpenClaw? 

 OpenClaw AI (originally known as Clawdbot or Moltbot) is an open-source, autonomous AI agent framework that represents a significant shift from "talking" AI to "acting" AI. Created by developer Peter Steinberger in late 2025, it went viral in early 2026 for its ability to operate a user's computer independently, execute tasks across different apps, and run 24/7.

Unlike a standard chatbot that waits for you to type a prompt, OpenClaw is designed to be a "digital employee" that lives on your hardware and communicates with you through messaging apps like WhatsApp, Telegram, or Discord.

Business Benefits

Technology needs to add value to business.  OpenClaw takes any instruction from any messaging app to execute the backend service seamlessly using AI actions.  It's game changer for the end user accessibility.

Four Core Technologies That Set It Apart

The primary difference lies in the transition from Large Language Models (LLMs)—which focus on generating text—to Agentic AI, which focuses on executing workflows.

1. The "Heartbeat" Mechanism

Most AI is dormant until you click "send." OpenClaw features a "heartbeat"—a background process that allows the AI to trigger itself based on time (cron jobs) or external events (like receiving an urgent email or a GitHub notification). This makes it feel like an assistant that is "always on. 

2. Semantic Snapshots & Computer Use

While early agents used slow, expensive screenshots to "see" a screen, OpenClaw often uses Semantic Snapshots. It parses the "Accessibility Tree" of a website or app (the text-based map used by screen readers) to understand where buttons are. This makes it faster, cheaper, and more accurate at navigating the web for you. 

3. "Soul" and Persistent Memory

OpenClaw uses a dual-memory system: JSONL Transcripts: A literal log of every action and word + Markdown Memory: A MEMORY.md file where it distills "lessons learned" 

4. Model Agnostic

OpenClaw is not a model itself; it is a gateway. You can plug in "brains" from different companies like Claude, GPT, etc. 

Stay tuned for more hands-on publications using Moltbot OpenClaw!

ML to AgenticAI

The landscape of Artificial Intelligence is moving fast, often leaving professionals wondering how all the "buzzwords" actually fit together. To leverage AI effectively, it is essential to understand that the ecosystem isn't a collection of separate tools, but a nested hierarchy of capabilities.

The Layers of Intelligence

1. The Foundation: AI & Machine Learning (ML) At the broadest level, AI encompasses everything from basic chatbots to robotics. Machine Learning sits within it, focusing on systems that learn from data to handle tasks like fraud detection and predictive analytics.

2. The Engine: Neural Networks & Deep Learning Moving deeper, Neural Networks mimic human brain structures to master pattern recognition. This evolves into Deep Learning, the powerhouse behind the Large Language Models (LLMs) and Small Language Models (SLMs) we interact with today.

3. The New Frontier: Generative & Agentic AI While Generative AI creates content, we are now entering the era of Agentic AI. Unlike standard models that wait for a prompt, Agentic AI is designed for autonomy. These systems don't just suggest a solution; they execute it through:

• Autonomous Agents: Operating independently to reach a goal.

• Personalized Automation: Tailoring workflows to specific user needs.

• Adaptive Systems: Learning and adjusting based on real-world feedback.

Why the Distinction Matters

Understanding this ecosystem allows businesses to move beyond simple "chat" interfaces and begin building goal-oriented systems. By transitioning from Generative AI to Agentic AI, organizations can shift from manual task management to intelligent, self-sustaining workflows.

The future isn't just about AI that talks; it’s about AI that acts.

25Years

Today, Hema and I celebrate 25 years together—a true milestone that demands a moment of pause, deep reflection, and joyful celebration. Looks like 10th anniversary was just the other day! https://photos.app.goo.gl/w4LsSzZgbuc4u5mU6  25 Years: A Beautiful Pause for Reflection. 

Over this quarter-century, we’ve moved across countries, faced countless challenges, and created a lifetime of shared memories. The constant that always stands out is this: through every "thick and thin," we have remained fundamentally united, standing side-by-side.

Instead, marriage is an active creation. It’s the continuous process of re-creation, nurturing, growing, lifting, loving, giving, and supporting each other—every single day. I genuinely believe that being a great leader in the world starts right within the four walls of our home.

You, Hema, is the core backend resilient service to enable the stability and success of my frontend visible career, publications and academic life.  I couldn't achieve with alternative - it's only by your sacrifice. 

Our relationship is a dynamic co-creation, and I am still a grateful student, learning how to make our partnership even stronger.

Hema, thank you for being the constant in this incredible ride. I am deeply grateful for everything we have built together, and excited for all that still lies ahead.

❤️ Happy 25th wedding anniversary, my dear. ❤️

ServiceNow Claude

ServiceNow is officially making Anthropic’s Claude the heartbeat of its platform. 

This isn’t just a tech upgrade; it’s an AI takeover. By deploying autonomous agents for clients like Panasonic Avionics, ServiceNow is moving beyond traditional SaaS and becoming the AI workflow backbone for complex, regulated industries.

For the market, the message is loud and clear: ServiceNow is no longer just managing your workflows—it’s automating them at scale

Press release at https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-and-Anthropic-partner-to-help-customers-build-AI-powered-applications-accelerate-time-to-value-and-apply-trusted-AI-to-critical-industries/default.aspx 

PS: On listening AI paradox recently at industry, proud to be part of my firm, to invest & inspire AI consistently in people/culture for last 14+ months.

AI Paradox

For years, enterprise software was the "safe bet" of Wall Street—predictable, subscription-based, and essential. But today, a "double-whammy" of disappointing outlooks from industry titans has sent the sector into a tailspin, as investors begin to fear that artificial intelligence isn't just a tool for these companies, but a potential replacement for them.

The tremors started with SAP, which saw its shares plunge more than 16%—its biggest daily loss in years—after issuing a 2026 revenue forecast that fell short of analyst expectations. The German giant’s cautious cloud outlook served as a warning light for the entire industry.

Closely following was ServiceNow, which dropped 11%. Despite reporting subscription revenue that actually beat estimates, the results weren't "convincing enough" to drown out the growing narrative that AI disruption might eventually shrink the market for traditional workflow tools.

For the first time in a decade, the "Software as a Service" (SaaS) model is being questioned. As AI agents and autonomous coding tools become more capable, the traditional software giants are racing to reinvent themselves through massive M&A deals and new AI-integrated features. 

20 AI daily tools

 

 Stop Working Harder, Start Working Smarter

We’ve officially moved past the "AI is coming" phase—AI is here, and it’s transforming how we handle everything from research to coding. If you feel like your to-do list is growing faster than you can keep up, it might be time to audit your toolkit.

The infographic below highlights 20 power players in the AI space to shift your workflow from manual to autopilot.

LLM steps

The rapid evolution of Large Language Models (LLMs) has transformed how we interact with technology. From writing code to offering medical insights, these systems feel almost human. But behind the "magic" lies a rigorous, multi-stage lifecycle. 

Based on the expert roadmap provided by Rocky Bhatia, here is a breakdown of how an LLM is built, refined, and maintained.

Stage 1: The Foundation – Data & Architecture

Before a model can "think," it must be fed. This initial phase focuses on gathering the raw materials of intelligence.

Stage 2: Training – Learning to Predict

With the architecture set, the model begins its "education."

Stage 3: Alignment – Fine-Tuning for Quality

A model that knows everything is useless if it isn't helpful or safe. This stage "aligns" the model with human values.

Stage 4: Deployment & Interaction

Once the model is accurate and safe, it is ready for the real world.

Stage 5: Evolution – Performance & Continuous Learning

The journey doesn't end at deployment. Language and information are always changing.

Enterprise AI strategy

 

This image perfectly captures the "AI Paradox" that many enterprise leaders are currently facing. While the board is excited about the shiny "AI Strategy" peak, the engineering teams are deep underwater dealing with the messy reality of technical debt.

Every executive team wants to "move fast" on AI. They see the summit—generative models, automated insights, and transformative growth. But beneath the surface of a sleek AI Strategy lies a massive, cold reality: the technical debt of the last twenty years.

If you try to build a high-speed AI engine on a foundation of "spaghetti" code and siloed data, the iceberg won't just slow you down—it will sink the project.

To move from the "Excited Executive" boat to the "Successful Implementation" submarine, teams need to stop treating AI as a separate project and start treating it as the ultimate reason to clean house.

• Audit Before You Automate: Document your legacy logic before asking an AI to refactor it.

• Data Hygiene is AI Strategy: If your data is messy, your AI will just be "fast and wrong."

• Empower the "Foundations" Team: Give the engineers in the middle boat the resources they need to fix the base. You can't reach the peak if the bottom is crumbling.

2026 Case Studies

Theoretical knowledge of databases and load balancers is a start but seeing how they are applied under extreme pressure is where the real learning happens. These case studies teach you about:

• Scalability: How to grow from 1,000 to 1 billion users.

• Availability: Ensuring services never go down, even during updates.

• Latency: Reducing the milliseconds between a user's click and the system's response.

1. YouTube & MySQL (The Power of Sharding)

Most people assume a giant like YouTube would abandon traditional relational databases for something "fancier." However, YouTube famously scaled MySQL by using Vitess, a database clustering system.

• The Lesson: You don't always need a NoSQL database to scale; horizontal sharding can allow traditional SQL databases to handle massive throughput.

2. Uber’s Geospatial Indexing

Uber processes over 1 million requests per second to match riders with drivers. This isn't just a simple query; it requires calculating distances in real-time.

• The Lesson: Uber uses hexagonal hierarchical spatial indexing (H3). By dividing the world into hexagons, they can quickly calculate "proximity" without taxing the CPU with complex geometry.

3. Meta’s Serverless Scale

Handling 11.5 million serverless function calls per second is an exercise in extreme orchestration. Meta (Facebook) has to manage resource allocation and "cold starts" at a scale most companies will never see.

• The Lesson: Serverless isn't just for small hobby projects; with the right infrastructure, it can power the world's largest social network.

System design is less about finding the "perfect" tool and more about understanding trade-offs. Whether it's choosing between the consistency of Google Docs or the massive throughput of Kafka, these 12 examples provide the blueprint for modern engineering.

AI Journey 2025 to 2026

As 2025 draws to a close, the AI landscape has shifted from "chatbots as a novelty" to "AI as an omnipresent teammate." We’ve moved beyond asking if AI can write an email to asking which model should run our entire department’s workflow.

Here is a year-end wrap-up of the "Big Three"—ChatGPT, Gemini, and Claude—and how they’ve redefined themselves in 2025.

Feature	ChatGPT (GPT-5.1)	Gemini (3 Pro)	Claude (4.5 Sonnet)
Best For	Daily Versatility	Multimodal/Big Docs	Coding & Nuanced Writing
Context Window	~128K - 1M (varies)	1M - 2M Tokens	200K - 1M (beta)
Key Innovation	Thinking Mode (Adaptive)	Computer Use/Native Video	Claude Code / Agentic Flow
Personality	Direct & Intuitive	Analytical & Integrated	Thoughtful & Cautious

Looking Toward 2026: The "Invisible" AI

As we flip the calendar, the trend is clear: AI is becoming ambient. We are moving away from dedicated apps and toward a world where AI quietly operates in the background—summarizing our meetings, organizing our files, and predicting our needs before we even type a prompt.

The bubble hasn't burst; it has simply soaked into the fabric of how we work.

Final Thought: In 2023, we were amazed it could talk. In 2024, we were amazed it could see. In 2025, we’ve finally started letting it do.

Happy New Year 2026 to all of you!

 

Code to Compost

As we close out 2025, I’ve been reflecting on the diverse paths our lives take. This week’s blog is a tribute to my UnderGrad classmate, Partha, whose journey over the last decade serves as a profound masterclass in purpose and resilience.

Life Transformation

After twenty years in the software industry, Partha made a choice many of us contemplate but few dare to execute: he traded the corporate office for the open field. Today, he is a seasoned natural farmer, leading the way in sustainable agriculture at the Bhoomi Network’s Farmer’s Center in Tiruvannamalai, India.

The Courage to Pivot

Spending time with Partha at his farm was a mesmerizing experience. Witnessing his transition from two decades of IT logic to the intuitive, rhythmic cycles of nature was deeply humbling. While many of us talk about "getting back to our roots," Partha has spent the last ten years living that reality—transforming rough, challenging land into a thriving ecosystem.

My visit left me with a few powerful realizations:

• The True Value of Nature: Seeing the interdependence of soil, water, and weather firsthand.
  
  
• Grit and Commitment: The physical and mental stamina required to build a farm from scratch is immense.
  
  
• A Student’s Mindset: Partha’s success stems from his passion for asking the right questions and his willingness to keep learning from the land.

A Hub for Ecological Change: The Vivasayi Maiyam

Partha’s work at the Bhoomi Network's Farmer’s Center (Vivasayi Maiyam) in Tamil Nadu is more than just farming; it is an educational movement. The center serves as a bridge for those looking to transition from conventional methods to "nature-imitated" practices through several structured programs:

1. Volunteering (7+ Days): A hands-on entry point for those over 18. Participants engage in farm work and group discussions under a unique "Gift Culture" model (suggested ₹500/day).
2. Village Immersion (3–5 Days): A program designed for urban professionals to experience rural social realities by living and working alongside village communities.
3. Holistic Internships (1 Month+): For those seeking a deeper, comprehensive understanding of ecological systems.
4. Model Farm Creation (6 Months+): A rigorous, experiential program where participants manage a one-acre plot, learning everything from seed banking to animal integration.

Final Thoughts

We often think of success in terms of vertical growth within an industry. Partha has shown me that success can also be horizontal expanding our impact across different sectors and aligning our work with the health of the planet.

I am incredibly proud of my friend. His journey reminds us that it is never too late to redefine who we are and what we contribute to the world.