Vision
I'm building an AI-native indie game studio. One human director, a crew of specialized AI agents (design, engineering, art, audio, narrative, QA), and a desktop platform that takes a serious indie designer from "interesting idea" to "commercial Steam title" — end-to-end, in any genre.
The wedge
Every department in a real studio is held together by something invisible: a coherent creative vision that survives the handoff from design to art to audio to engineering. The reason small teams struggle to scale and solo developers struggle to ship is that holding that vision across every department is the actual hard part — not writing the code or generating the assets.
The platform's wedge is a design intent graph: a single canonical artifact that holds design intent and standards across visual, narrative, and audio — style bibles, character profiles, lore schemas, audio palettes — at every tier of granularity. Director-level direction ("more dramatic, slower pace") and node-level direction ("set this property to that value") mutate the same graph through the same API. Promotion and demotion between tiers is a UX affordance, not a data conversion.
Around the intent graph hang the things that make it useful: an asset and lore pipeline that turns intent into realized content, an intent→Godot translator that takes a design and produces a working project, and a build/ship pipeline that puts the result on Steam.
What v1 looks like
The bar for v1 is "idea to functioning complex 2D game in any genre," delivered end-to-end through the platform alone. Four to six months. No best-in-class checkbox per component — just the minimum that works for v1, then enhance once it's shipping.
Load-bearing for v1:
- The design intent graph.
- Intent→Godot translation.
- An asset/lore pipeline minimal enough to take design intent and emit usable content.
- A studio-agent runtime that composes departments under human direction.
Deliberately skinny or deferred for v1: SFX generation, deep 2D asset chains, in-game UI drawing, runtime LLM dialogue, accessibility implementation, launch and live-ops. Those enhance post-v1, on top of the same wedge.
How it gets built
In the open. Real Claude Code session transcripts where the work actually happens — including the recalibrations and the dead ends. The progress matrix on the home page tracks every platform component from research to 1.0, and updates the day work lands. RSS and email get the same posts on the same day. No funnel, no paywalled tier, no engagement-bait.
Where it goes
Past v1, the post-v1 enhancements stack on top of the same wedge: deeper art and audio pipelines, multi-agent execution within departments where the harness shows it pays off, post-launch live-ops support, BYO-API-key flexibility for power users.
Long-term, I want this to be the place where a serious indie can iterate on a game with leverage they couldn't otherwise have — without giving up authorship, taste, or the right to ship the thing they actually wanted to make.
Who I am
Scott Florida. Reach me at scott.d.florida@gmail.com.
The name
"Project Daedalus" is the working codename — a nod to the mythological craftsman who built things that worked. It may not survive to launch (people aren't sure how to pronounce it on sight, which is a real friction for a consumer-facing brand), so the public name may shift before beta. The devlog is the canonical home of the project either way; the URL and brand string are the only things that move.