Margaret and I left our jobs in February 2025. We’d both spent years shipping software at companies that were good at it — Hinge Health, HubSpot, eSentire — and we’d each developed a pretty clear picture of what makes a team fast and what makes a team slow. We agreed on most of it. So we decided to build a company together.

This is the story of why we started okthink, and the two ideas that drive everything we do.

Get things done, then make them pretty

The first idea is about operational efficiency. Not the kind you put on a slide deck. The kind where you look at how work actually moves through your day and ask: is this the fastest way to get from idea to shipped product?

Most engineering orgs obsess over architecture diagrams, design documents, and review processes. Those things have their place. But we’ve seen too many teams spend weeks debating the perfect abstraction while a competitor ships the ugly version and learns from real users. The architecture doesn’t matter if the product never reaches anyone.

We care about the loop. How fast can you go from “I think this should exist” to “it’s in someone’s hands”? Every tool we adopt, every process we create, every decision we make — we evaluate it by whether it makes that loop faster or slower.

This mentality turned out to be the perfect foundation for the AI era. When agentic coding tools started getting good, we didn’t see them as a novelty. We saw them as the biggest operational improvement we’d ever encountered. We’d already been asking “how do we ship faster?” for years. AI agents were the answer we’d been waiting for.

Today, we orchestrate multiple AI coding agents in parallel. We built our own tooling to manage them. A two-person team producing the output of a much larger one — not because we work longer hours, but because we spent years thinking about how to eliminate friction from the development process. AI just removed the last bottleneck.

Be a duck

The second idea is about what kind of people should build software.

A duck can swim, walk, and fly. It’s not the best swimmer, the best walker, or the best flyer. But it can do all three, and that makes it one of the most adaptable creatures on the planet.

We think people who build technology should be ducks.

The industry has spent decades splitting “product,” “design,” and “engineering” into separate disciplines with separate career ladders, separate meetings, and separate ways of thinking. We get why it happened — specialization scales. But it also creates seams. Every handoff between a designer and an engineer, every misalignment between a product manager’s vision and what actually gets built — that’s a seam. And seams are where quality breaks down.

Here’s what we believe: designing a product, coding a product, and envisioning a product are not three separate games. They’re three facets of one game. The best software comes from people who can hold all three in their head at once. Someone who can look at a screen and see the user experience, the technical constraints, and the business opportunity at the same time — that person ships better work than any three specialists handing documents back and forth.

That doesn’t mean everyone has to be great at everything. The duck isn’t the best at any one thing. But it never gets stuck. It can always make progress because it always has another mode of movement available.

At okthink, both of us design, both of us code, and both of us think about product. We don’t have handoffs because there’s nobody to hand off to. That’s not a limitation. That’s the advantage.

That’s also why the duck is our mascot. It’s not a random choice. It’s the whole philosophy in one animal.

What we’re building

Our first product is Itsumo — an AI-powered study app for the Japanese Language Proficiency Test. It’s the only app that covers all five skills the test measures. We started there because Raf has studied Japanese for a decade and was tired of juggling five different apps just to prepare for one test.

But okthink isn’t a language app company. It’s a platform for building AI-powered tools that help people develop real skills. Itsumo is the first product. Not the last.

Why this blog exists

We’re going to write about what we learn as we build. The tools we use, the decisions we make, the things that work, and the things that don’t. If you’re interested in AI-native development, agentic workflows, or just want to watch two engineers try to build something worth building — stick around.

Welcome to okthink.