In July 2025 I founded a game studio believing that a small senior team running AI woven into every discipline could produce at the scale of a much larger organization. This page is what happened, including the ending, which is where the data led.
Rift Fighters, a deck-driven, real-time cinematic fighting game built in Unreal Engine. Think Nemesis-System relationships meets card strategy meets action-movie spectacle. In one self-funded year, a five-person senior core produced a UE5 cinematic vision video (directed, shot, and cut in Sequencer), four playable prototypes built with AI-assisted engineering, a complete game design document, and an evidence-based financial model that ultimately decided the company's fate. The video premiered at a Seattle Indies Founders Club session where Rift Fighters was the featured project, hosted by Harebrained Schemes co-founder Mitch Gitelman.
The core team I recruited: Roy McCormack (designer of the Nemesis System), Joel Blakely (Lead Character Artist on Shadow of Mordor), Branden Conley (Lead Gameplay Engineer, six years on Magic: The Gathering Arena), and Dan Burdick (core-game designer on Magic: The Gathering).
Around that core I ran a specialist bench, bringing people in as the work demanded it: backend engineering, technical art, UI/UX, monetization design, data analysis, financial modeling, and marketing. Seven disciplines, seven people, inside twelve months.
Advising us: Mitch Gitelman (co-founder, Harebrained Schemes), with informal counsel from Dave Luehmann (GM of four Microsoft game studios; Director of Game Development at Netflix) and a deal advisor with two major studio acquisitions behind him.
This is the Portfolio Cut — the vision video with protected design elements removed, cut to music. The complete cut, with dialogue and interface intact, is the Master Cut. Ask to see it →
TIMELINE: JUL 2025 — JUL 2026 · CLIP PLACEMENT APPROXIMATE
The structure was a small senior core, a specialist bench called in by discipline, and AI connecting every discipline to the rest.
The multiplier wasn't any single tool. It was treating AI as infrastructure with a memory, not a chat window, then holding every output to production standards. By March 2026 the studio's living use-case log recorded more than twenty distinct AI use cases across eight departments, with average time savings around 60%.
Claude connected to our ClickUp docs and Google Drive, plus a shared library of markdown files that served as the company's persistent brain. Decisions, design rationale, and pipeline knowledge stayed queryable instead of trapped in threads. Fully wired together May–July 2026.
AI-assisted Python tools for Maya (animation fixes and a WYSIWYG slow-motion authoring workflow) built in days instead of sprints, by an artist, without waiting on an engineering queue.
Four playable prototypes stood up with AI-assisted engineering, letting design questions get answered by play instead of debate, backed by an automated tuning harness that ran thousands of simulated fights.
AI-assisted deep research, generative image and video for pre-vis, and music production feeding the vision video and pitch materials. One person covering ground that used to take a department.
The model said the game can't earn back what it would cost to build. So I believed my own instrument.
I rebuilt our financial model from the ground up: 21 formula-driven tabs, more than twenty versions, every load-bearing assumption validated against 15+ sourced research briefs. I gut-checked it against genre reality, and the answer came back unambiguous. On July 21, 2026, I put the conclusion in writing to my team and advisors and wound down development. The methodology outlived the game. Proving a small senior team can produce at studio scale, then stopping when the numbers said stop, is what I bring to whatever I build next.