In 2014, I put my first real estate agent on a roller coaster in a DK1. She was 50 feet up in VR, looking down at a construction site that didn't exist yet. Her knees buckled. She got it.
That became the pattern: build it, ship it to the field, watch what breaks. Two decades, 53 countries, and counting.
This isn't a portfolio. It's a field manual for dragging prototypes into the real world before they're ready—because that's the only way to know if they're worth finishing.
The Method
No timeline. No résumé. Just the patterns that survived two decades in the field.
Nobody funded VR until they felt the plank. Nobody adopted Matterport until they saw the 95% enquiry lift. If you can't demo it in 90 seconds, you don't understand it yet.
The lesson: Visceral beats logical every time.
Picked up VR in 2014, AR in 2016, jumped to GraphQL pipelines in 2018, LLMs in 2022. Not chasing trends—watching where the infrastructure is about to get good enough to ship.
The lesson: Infrastructure timing matters more than being first.
Conference circuit
Traditional consulting: fly in, sit in conference rooms, produce recommendations, leave. Six months later, implementation still pending. We work directly with teams to build, test, and deploy. We're not selling advice. We're shipping code.
The lesson: Functional prototypes beat PowerPoints.
You can't learn to swim by reading about swimming. People need to understand what these systems can and can't do. Quick validation sprints, exploratory phases, hands-on tools like LLM Adventure—you learn by doing, experimenting, failing safely.
The lesson: Intuition comes from experience, not presentations.
Learning by doing
Field Reports
Narrative vignettes. Each one a mini-story with conflict, resolution, and a lesson that stuck.
2014–2016
Started with roller coasters in 2014. Put agents in DK1s on virtual coasters—same effect as the plank would be later. They'd grip their chairs, knees wobbling, completely sold on VR before we showed them a single property.
At a 2015 hack day, we built the plank—literally nailed boards together. Made people walk it while wearing the headset. Nobody forgot after that. It became our traveling torture device. AREC, SXSW, internal roadshows—anywhere skeptics gathered. By 2016, we'd strapped hundreds of people to it. Real estate agents fought to go next.
The lesson:
Visceral beats logical every time. Data gets budget. Fear gets attention.
2015
Matterport cameras were $4,500 and took 45 minutes per property. Real estate agents didn't care about the tech—they wanted to know: Does this sell houses?
We built the ingestion pipeline first—capture, QA, publish same-day—then tracked everything. Listings with 3D tours got 95% more email enquiries and 140% more phone reveals. The data convinced the board. The tours convinced the agents.
The lesson:
Pipeline work is invisible until you measure it. Then it's the only thing that matters.
2016
Brought Melbourne's warehouse-scale VR startup into our AREC booth. Let agents shoot zombies in free-roam VR before showing them property tours.
Why? Because after adrenaline, calm feels profound. They'd take the headset off, look at a normal listing tour, and say: "Wait, I can actually use that."
The lesson:
Context shapes perception. Sometimes you need to break someone's expectations before you can reset them.
2016
Worked with Plattar to build AR brochures for Ray White. Dozens of image activators—point your phone at the page, 3D overlays pop out. Floor plans, neighborhood maps, building renders.
It was 2016. Nobody was doing this yet. We printed 500 brochures, distributed them at AREC, and watched agents film each other discovering it.
The lesson:
Sometimes you just need to show up first. The tech doesn't have to be perfect—it has to be there.
2019
A European automotive client lost their white-labeled vendor mid-contract and needed a replacement capture pipeline in 4 months or they'd lose the contract.
We rebuilt the entire AR360 stack—Android/iOS capture apps, Unity scene stitching, and a desktop QA suite—to keep their inventory live. Barely made it. Contract saved.
The lesson:
Constraints clarify. When there's no time to overthink, you ship what works.
Residency stack
2023–Present
Embedded 4–6 week builds with 30+ teams. Deliver RAG stacks, teach prompt design via a text adventure game, leave them with working Pinecone/Chroma infrastructure.
Same operating system as REALABS: show up, build the thing, hand over the scaffolding.
The lesson:
The work hasn't changed. The raw materials have.
The Lab Bench
Not an archive. An active workshop. Build logs, decision rationale, and shipped code.
Consolidating ingestion for LLM residencies. GraphQL schemas evolved into RAG plumbing.
What happened to GraphQL360. The capture stack that became an ingestion pipeline.
Mobile 360° capture platform. Built for photographers, evolved into Twirrl.
First property portal in VR. Google Daydream launch. Sticky-note walls → product decisions.
How to run a company from trains. Estonia OÜ backbone + async rituals that work.
Text adventure game teaching prompt engineering. Used in residencies to build evaluation muscle.
Work with us
Bring the field lab playbook to your team. Build working prototypes, ship to the field, watch what breaks.
The full residency experience. We embed with your team, build the thing, and leave you with the scaffolding.
Need a second opinion on your AI roadmap? 2-hour consultation, remote or on-site.
Far Horizons OÜ · Estonia · Available globally