A personal slice of the internet for AI tools, tips & tricks — and long-form LLM investigations.

Hey, I'm Manish Shahi — software engineer by trade, AI dabbler by obsession, and technical writer whenever something finally works and I refuse to keep it to myself.

Welcome to my little corner of the internet. I try to cut through the AI noise and write about what's actually useful in developer tooling — the kind of notes you'd send a coworker, not a LinkedIn thread.

I've spent 7+ years building web applications, APIs, and AI-powered tools. My stack spans PHP, Node.js, Golang, and Python — languages that have all, at some point, broken my patience and saved my job. That engineering mess keeps the writing honest.

By day, I'm a software engineer at Inncelerator, shipping things real people actually click. After hours (read: also during hours), I'm poking at new LLMs, poking around MCP servers, and automating anything that annoys me twice.

Some days I'm stress-testing a new model until it politely gives up. Other days I'm writing the guide I wish had existed before I spent three hours discovering a one-line fix. I like building things — and I like explaining them without making it sound like rocket science.

This blog is me learning in public, so you don't have to make all the same mistakes. If you want less hype and more "here's what worked when I tried it," you're in the right place.

Lately that means the Inside LLMs series: I interview a model about its own architecture and behaviour, treat the chat like a first draft, then check every bold claim against arXiv, docs, and whatever else is public. The Kimi and DeepSeek posts are the template — diagrams, interactive labs, glossary links, and a downloadable transcript when it helps.

Between those deep-dives I write the shorter stuff too — MCP servers in production (PHP and otherwise), agent purchase flows, prompt and tooling notes — always from the angle of someone who still ships web apps for a living.

Got a tool that confused you, or a tip that actually worked? Let's talk — I read everything, even the chaotic drafts.