about me.
My name is Fred Abrahamson. I'm a designer, developer and ex-founder with 10+ years of experience shipping products end-to-end in fast-paced, uncertain startup environments.
I've built and experimented with technology from an early age, which led me to an education at Hyper Island[1]. Afterwards I co-founded, built and sold a company. It was the early days of chat interfaces (Slack and Siri were new), and our first experiment was a chat-based employee survey with an ML backend that generated follow-up questions. That experiment eventually became an automation tool for HR[2].
A hands-on education involving design, coding, workshops and more. We learned by applying it on assignments for companies.
Typelane helps HR reps give new employees a personal, engaging onboarding, and automates the tasks and coordination around it. It was acquired at the end of 2022.
- Team.I led the product and development team. I also handled recruiting and team development.
- Design.I was the only designer, so I worked closely with users and customers on their problems. I built quick prototypes, ran customer workshops and tested with users.
- Development.I worked on the frontend of our product and our internal tools.
Wanting to return to data work, I joined a Copenhagen-based data company as their interim UX lead[3] for a year to redesign their platform from the ground up.
GameAnalytics makes analytics for mobile game developers. Its users range from small indie devs to publishers with a billion downloads.
- Design.I organised, designed and documented a design system. It made the product more consistent and sped up design and frontend work. I took features from discovery and research to final designs and specs, and interviewed our users on a regular schedule.
- Development.I prototyped features in Figma and React to test ideas with the team and with users. Some were data visualisations built on real data.
- Product.On several projects I was the product lead for a cross-functional team of data scientists, engineers, frontend developers and marketers.
- Team.I mentored the junior and mid-level designers.
Afterwards I became part of a new company called Trail/Pley[4] as their product design lead. They had an ambitious vision to make mobile developers' games instantly playable in any web browser.
Pley is a gametech startup. Its technology makes any Unity game playable in any web browser, with no download or install. It was acquired by Xsolla.
- Design.I designed the web player, where people load games in their browser, pay, connect accounts and more. I also designed the Unity SDK, the developer console, and UA features that helped marketing teams move their campaigns from mobile to web.
- Product.I led product work on many of the features we shipped. With analytics, I A/B tested versions of the games’ website template and raised conversion at several points.
- Developer relations.I onboarded and supported customers as they got started on our technology stack.
I followed AI developments closely at Pley. When the company was acquired, I didn't want the next traditional design role. I wanted to dig into the tools and technology instead. Design was changing along with everything else, and I wanted to get ahead of it. I didn't take the hype at face value, so I set out to learn how these systems work first hand.
I started helping companies with AI rollouts and enablement, and built agentic systems into their internal processes and customer-facing applications.
My work with teams, the experiments I ran and what I learned from them accumulated into a passion project called Midland[5], soon to be open source. It's a shared context layer with a data model that teams can use to structure their company knowledge and distribute capabilities from one surface to any tool or team.
It started as a shared memory and skill library for one small team. It grew into a typed but flexible data model that agents and people can both understand, use and shape.
- Design for agents and people.I designed the MCP, the UI and the typed data model underneath. Agents and people read and shape the same content.
- Test harnesses.I built test harnesses that catch skill drift and measure hallucination rates, token use and more.
- Self-improving.When an agent gets something wrong, it logs the friction and the likely cause. Fixes to prompts and skills start from those frictions, so the next run goes better.
- Build with agents.I built it with Claude Code and my own agentic workflow. Agents triage, scope and build the work. I decide at each step.
- Open source.I plan to release it as open source, with a harness and skills, so anyone can adapt it with agents.
I aim to open source it rather than found a new company. I want to keep getting better at deploying these capabilities, finding and solving problems for people rather than racing the market with a new venture. I'd rather join one in its early stages.
If you'd like to discuss harness and context engineering, AI rollouts and enablement, or what design can become as agents and people work together, reach out to me here. I'm happy to talk for hours about it.