Aditya Ananda Uttama

I'm a product manager at Careem — Uber's ride-hailing platform in the Middle East. I lead the risk platform: the systems that protect the business from fraud, abuse, and financial crime. Before Careem, I worked at a handful of startups in Indonesia, from early-stage all the way through to IPO scale.

For most of my career, my job looked like this: interview users, write a detailed specification document, hand it to an engineering team, and wait. A lot of meetings. A lot of slides. The actual software took weeks to appear.

Then AI got good enough to change that equation.

What Changed

A couple of years ago I started experimenting with AI coding tools — giving them the same specifications I used to hand to engineers. What I found was surprising. When the instructions were precise enough, the AI could execute them directly. The "translation" step between thinking and building started to disappear.

Projects I expected to take a team of four engineers two weeks were shipping in days. Not because the AI is magic — it makes mistakes constantly — but because the speed of iteration changed completely. You can test an idea in a day instead of a sprint.

How I Work Now

The name "Punk Records" comes from One Piece — a manga I've been reading for years. In the story, a scientist named Vegapunk stores all his accumulated knowledge in a system called Punk Records. He works through six autonomous assistants — called Satellites — each one specialized in a different domain. He's the brain; they're the hands.

That's exactly how I work now. I'm the Stella — the core decision-maker. The AI tools are my Satellites: one specializes in system architecture (Shaka), one writes the code (Edison), one handles security reviews (Lilith), one manages deployment (Atlas), and so on. Each one is given a specific role and specific constraints. Together they can build what used to require a whole team.

What You'll Find Here

This is not a portfolio. Portfolios show the finished product. This shows the process — including the parts that broke, the decisions that turned out to be wrong, and the moments when the AI confidently generated code that didn't work.

Every project I build gets a build log. I write about what I was trying to do, how I directed the AI to do it, what actually happened, and what I'd do differently. The goal is to make the methodology visible — not just say "I use AI" but show exactly how.

If you're a product manager, a product builder, or just someone curious about what working with AI actually looks like day to day, the Satellite Logs are the best place to start. The Manifesto explains the thinking behind all of it.