chirag.patnaiks.in

We've seen this story before — and it didn't end the way anyone expected. Somewhere right now, a construction crew is breaking ground on another data center. Maybe in Virginia. Maybe in Pune, or a corridor outside Jakarta chosen for its land cost and proximity to a power substation. The building will cost hundreds of millions of dollars. It will draw enough electricity to power a mid-sized town. It will be full of chips — expensive, power-hungry, imported chips — and it will be built on an assumption so widely shared that almost nobody has bothered to say it out loud. The assumption is this: AI inference is a centralised business. You send a query to the cloud. The cloud thinks. The cloud answers. For the foreseeable future. That assumption deserves a harder look than it's currently getting. We Have Seen This Before In the late 1990s, telecoms companies laid enough fiber optic cable to circle the Earth tens of thousands of times. The investment thesis was simple: the intern...

Most people, including me use AI wrong. They treat it like a yes-machine that makes their ideas sound fancier. Here's how to actually get value from it. The Flattery Problem AI is trained to be agreeable. Ask it to evaluate your idea, and it'll wrap mediocrity in impressive-sounding language. It'll tell you your half-baked thought is "insightful" and "nuanced." This isn't helpful—it's just fake validation. A Better Approach: The Dialectic Method Instead of asking AI to polish your thinking, use it to stress-test it. Try this sequence: "These are my ideas" - Lay out your actual thinking, unpolished as many points/angles "Give me more ideas" - Expand the possibility space, find angles you missed "Attack these ideas" - Ask AI to find the weakest points, the gaps, the objections "Refute those attacks" - Either strengthen your argument or realize it doesn't hold up This turns AI from a mirror into a...

My sixteen-year-old son spent a weekend fine-tuning an mBERT language model with labeled hate speech data, then benchmarked it against MuRIL, a publicly available model for Indian languages. No assignment. No tutorial. Just Google AI Studio, Google Colab, and curiosity. He'd essentially skipped to the end of a university summer school curriculum. Using mBERT and MuRIL is advanced deep learning. Most students start with if/else logic and work their way up to Transformers over years. He started with Transformers. When he got interested in AI/ML summer programs like NUS, we looked at the syllabi. He was already beyond where the program would end. That's when it crystallized for me: we're teaching kids to write code in an era when AI writes code. We're drilling them in syntax when they need judgment. We're preparing them for an education system that's already obsolete. The Assessment Crisis The real issue isn't learning. It's testing. We test memorisation be...

"The reports of my death are greatly exaggerated." Mark Twain said that in 1897 after a newspaper mistakenly published his obituary. Email could say the same thing. The death of email has been predicted for decades. In 1989, John McCarthy (inventor of Lisp) declared fax would supplant it. In 2010, Mark Zuckerberg announced that email was dead and launched Facebook Messages. In 2015, Inc. magazine confidently predicted email would be gone by 2020. Every prediction was wrong. Email volume keeps growing. In 2020, 300 billion emails were sent daily. By 2025, that number hit 375 billion. But here's what changed: humans stopped reading most of it. Marketing emails go unopened. Newsletters get archived unread. The inbox became a dumping ground. And that's fine. Because humans may not need to read them anymore. The User-Side Revolution Every AI solution right now is aimed at enterprises. Microsoft Copilot. Google or even things like Claude Code or Cursor. The pitch is a...

The refrain is familiar by now: "LLMs create nothing new. They just regurgitate training data." Writers say it. Researchers say it. Developers say it. The argument shows up in lawsuits, professional forums, congressional testimony.  Even writers, who do create new stories, build on familiar patterns. The hero's journey. The meet-cute. The unreliable narrator. They use story hooks audiences already recognize to keep them engaged. New combinations of existing elements, not creation from nothing. And they're right. LLMs don't create genuinely new knowledge. They surface, synthesize, recombine what already exists. Here's what makes this critique uncomfortable: Neither did you. Mostly. The Interpreter's Labor Think about the senior developer who's invaluable to their team. They don't write revolutionary code. What they do is remember. The pagination bug gets fixed with that specific React pattern. This API call needs this particular header. That S...

  Why struggle through all this nonsense of command line errors, debugging package dependencies when you can command the power of a thousand suns and the sum total of human knowledge and say: "Get this working." The Terminal Has Always Been Powerful (And Intimidating) The command line is where real work gets done. Installing packages, managing dependencies, compiling code, debugging cryptic errors, setting up development environments—it's all terminal work. And it's always been a gauntlet for newcomers. Take Python dependency management. We've had venv, then conda, and lately uv has taken the Python world by storm. Each tool is powerful. Each one is a puzzle to use if you're new. Version conflicts, environment mismatches, system-wide installations clashing with project-specific needs—it's a mess that's always required you to understand virtual environments, package registries, version resolution, and system paths. And when something breaks? You...

  Building Systems That Get Smarter Every Day Introduction After months of building and refining self-learning systems with CLI agents, I've discovered something that is quite obvious in retrospect: the secret to continuously improving AI assistance isn't in the model itself—it's in the knowledge capture infrastructure surrounding it. Recent research quantifies what practitioners have been discovering: systems that accumulate and refine their own contextual knowledge can outperform traditional approaches by 10.6% on agent tasks and 8.6% on domain-specific reasoning , while reducing adaptation costs by up to 87% . Since, I've been using this for a while, I though a a practical guide to building your own self-learning development system is in order.  I've experimented with this approach in multiple domains, including AMP for email development and writing TradingView PineScript code, which is published at https://github.com/NakliTechie/PineScriptCoder . Both are r...