Cutting an 18-month plan down to three months
The full game needed a team of eight. We were less than one. Four decisions in one day: a 10-minute demo, AI-only art, first-person only, read-only blockchain.
Written after the fact, from the commit history and the design logs.
Today the project did the arithmetic it had been avoiding.
The 18-month plan on paper — stealth, open world, life sim, multiplayer, blue team, mission editor — was estimated at 7.75 full-time people. Available capacity was about half of one. So the plan became a minimal playable demo: ten minutes, one apartment, a hacking terminal, one or two CTF missions, backend integration, and a Dr1ftS0und UI with placeholder music. Everything else moved to the post-MVP backlog. The success metric written down that day: players want more, not “this is complete”.
Three more decisions landed the same day:
- AI-only art. No budget for artists, so all 2D art through a local pipeline: ComfyUI, SDXL, and a custom h4shDr1f7 LoRA to be trained for consistency. Fallback if it failed: asset store placeholders.
- First-person only. The Drifter is a pair of hands at a keyboard. No body, no mirrors, no customization. A full character was twenty to forty hours the project did not have.
- Read-only blockchain. The client could query the chain, never write to it. No gas fees, no wallet friction for demo players, mission results in Postgres.
The same week’s asset audit counted 95 models, 50 textures and zero audio in the project.
The spoiler, from a year later: none of that plan’s ship gates were ever met. But cutting to ten minutes is the reason anything playable existed at all, and two of the four decisions still stand. The Drifter is still never shown. And the demo still touches no chain at all.