Articles

Tokens, embeddings, parameters, pretraining, fine-tuning, reinforcement learning, inference, temperature, RAG, hallucinations — all on one question, from the keystroke to 'Paris'. A builder's notes after going back to the fundamentals.

Shreyas Doshi is right: AI won't save you from building the wrong thing. I have the graveyard to prove it — a year of systems my agents built fast and correctly that one cheap test would have killed before any code. Here's the test, and where judgment actually comes from.

I once improved a mechanic's first step by 30% and the loop still didn't close. Here's what that taught me about borrowing what works elsewhere — and why it matters more now that building is nearly free.

How I decide whether an AI agent is allowed to act on its own — the evaluation gates, the three products I killed with the numbers shown, and the outage that changed how I monitor everything. Written for product people. No jargon, no returns, no hand-waving.

Replit's 'self-driving company' essay shows what agent fleets can do once trusted. This is the architecture underneath such a system — seven layers, drawn from one I operate in production, held together by two rules: no agent promotes itself, and no agent grants itself resources.

Grading a chatbot's answer is one problem. Deciding whether to let an agent touch production is a different one. You don't score the output — you grant authority, in stages, behind gates that it has to earn.

I built and ran a working automation without opening the tool's UI — driven end-to-end from Claude Code through MCP. The interesting part isn't that it worked. It's that the build path is itself guardrailed.

A copilot suggests; an agent acts. The moment an agent can act on its own, one question becomes the entire product: is it confident AND safe enough to do this without a human? Here's a small, working pattern — with the actual workflow.

An evaluator that catches a bad output after the fact is sometimes too late — you can't un-send a webhook or un-drain a budget. RFC #7218 proposes the missing layer: a deterministic, pre-action safety gate for autonomous agents. Here's the design.

The hardest part of building with AI agents isn't getting them to work. It's noticing when they've quietly stopped. Five reliability lessons from a year of running autonomous agents that trade real money.

From 250 orders a day to 3.5x users and 2.8x revenue in 18 months. Here's the full playbook — discovery, ops, acquisition, and AI/ML — from inside the building.

Introduction: In the era of outcome-oriented models driven by AI and ML technologies, the landscape of UI/UX design is undergoing a profound transformation. As products and industries embrace declarative,...

Achieving 3x monthly run rate in user and revenue growth in just 18 months! 📈 Rarely shared publicly, but here's the behind-the-scenes story of our remarkable journey. ✨ Market Insight: - Kuwait's population: 4.82...

Innovative strategies are crucial for maintaining a competitive edge, particularly in user engagement and activation. This post outlines our integration of Artificial Intelligence (AI) and Machine Learning (ML) to...

Introduction - Wanted to share the story of working with Mashkor for the initial 8 months working here about how we worked to...

Wanted to share some learnings about MVP's. In our fast-paced world of product and growth, especially for domain in our which is highly operation-driven business, the mantra was clear to test out ideas soon – MVPs...

Having received numerous inquiries from my network about the nuances between Growth Product leaders and traditional Product leaders, I thought to shed light on this topic from all the work I have been doing and how its...

In the 8-week boot camp project with GrowthX, after the four weeks of deep diving in modules, we had to apply the concepts and fundamentals we learned to real-time products. We got the...

I have been a Digital Marketing professional for eight years and specialized in SEO for seven years. I have worked with numerous varied projects to achieve results. I also have run websites for fun and flipped them...

My first surfing lesson. You cant stop the waves, but you can learn to surf. It took me 90 minutes to get it manageable — crumbled a few times, but got there.

Part-1 (Jan-18) Lately, I have been continually asked for inputs from a lot of friends, for the quick physical transformation in a short time. I have been...

This is an idea has had a huge impact on me, a long time back unknowingly and now knowingly. I will try to discuss only the emotional pain, causes and some remedies to deal with it. Also, in many cases, even physical...

I have been intuitive and impulsive for most of early life, which has got me into trouble on a lot of occasions. I have been lucky to receive tough love several times from special people in my life and realize some of...

Happen to just come across this image, and it just hit me to the core. I just kept reading about Empathy and Opinions in so many places. It really did not make me understand to clearly, a lot of my Mentors and the...

India ranks 130th amongst the 188 countries on the Human Development Index, however, once inequality is factored in, we drop even further to rank 150th. In addition to employment strategies, the report calls for the...

Howdy, So my last vacation of my year was in Thailand, I just loved the vibes of that country, on a personal experience it had a blend of the feeling you get in India an Island as well. The moment I landed in...