Hi, I’m . I teach machines to work, and design for the people beside them.

Now founding a company in . Before that, I led Infrastructure and Foundations at , Product at and founded in my dorm room at 23, while studying computer science at .

Copenhagen00:00:0055.68° N12.57° E

(02) Writing

23 Oct 20266 min readSuperintelligence is a storage problemMost of the money and talent in AI is going into making models smarter, which makes sense. Much less is going into where an agent keeps what it knows between one model call and the next. I think that second part decides more than it looks like, so this is an attempt to explain why, with some math.A model on its own remembers nothing between calls. Everything that makes an agent useful over weeks lives outside it. The conversation so far, the facts it has learned about the people it works for, the files on the computer it works on, and the run itself, meaning which steps are done and which are waiting on a person or a timer. On top of that sits a record of every model call, which you need later to explain why the agent did what it did.I believe in the Bitter Lesson, so I assume the model under all of this gets replaced by a better one every few months. That only works if the state survives the swap. It also has to survive a crash halfway through a step, it has to cost very little while the agent sits idle, which is most of the time, and someone has to be able to delete it properly.

8 Oct 20263 min readWe built the crew before the starshipIn 1948 two law partners in Milwaukee, Elmer Winter and Aaron Scheinfeld, had to bring in temporary help to meet a deadline. The company they started to send people into offices became Manpower, one of the largest staffing companies in the world. What they sold needed no integration. It was a person, and every office already knew what to do with one.78 years later the crew is ready. Models write, reason and code at a level few of us expected this soon, and they pass the Turing test in writing, on video and in voice. Yet most people still use them as a smarter search box. The bottleneck has moved from intelligence to distribution.In January 2007 none of the parts inside the first iPhone were new. Putting them in one thing that just worked was the invention. The parts of a colleague are on the table now. A voice that sounds like a person, a computer of its own, an inbox, a phone number, a calendar, a wallet and a memory that outlasts every model. Nobody has put them in one body.

2 Oct 20261 min readLight through paperI’ve always wanted a gradient builder that makes gradients as beautiful as OpenAI’s. The ones on its old plugin cards: soft washes of cornflower and mint, one edge almost in focus, light leaking in from a corner.For two years I tried to build one from Justin Jay Wang’s methods for random gradients. With Claude models, again and again. Every time: close, and wrong.Last week I tried again with Fable 5.1. It blew my mind. Every card on this site sits on what came out. Try it.

23 Oct 2026Superintelligence is a storage problemWhere an agent should keep what it knows
8 Oct 2026DraftWe built the crew before the starshipNow somebody has to build the Enterprise
2 Oct 2026Light through paperGradients as beautiful as OpenAI’s

Hi, I’m . I teach machines to work, and design for the people beside them.

(02) Writing

About Oscar Antonio Borlund Orellana

Work

Stealth, Founder, Jun 2026 – Now

Legora, Engineering Manager, Jan 2026 – Jun 2026

Beyond Work, Product, Oct 2024 – Dec 2025

Corti, Senior Software Engineer, Sep 2023 – Sep 2024

Rig.dev, Founder & CEO, Jan 2021 – Aug 2023

Aarhus University, BSc and MSc Computer Science, Sep 2018 – Jan 2023

Hi, I’m Oscar. I teach machines to work, and design for the people beside them.

(02) Writing

Hi, I’m . I teach machines to work, and design for the people beside them.