▸ Examples

Three pipelines. All shipping.

Symbiotica is too young to have a long case-study list. Here's what's real today — what's shipped, what's running, and what we use ourselves.

IMPERIA ONLINE·2026 →

Bakery Story — full event-asset pipeline

Imperia produces themed live-ops events for Storm8's long-running mobile game Bakery Story. Each event ships a curated bundle of 4 to 28 assets — themed ovens, food, decorations, viral coin variants — to a strict spec (slot count, naming, plot grids, rotations).

We built the end-to-end pipeline. A game designer drops a brief into Telegram (`/idea`). A Planning agent triages it into a clean backlog card. The Designer agent takes the brief, proposes three theme directions, locks one, expands it into the narrative hook, and emits the 28-asset prompt list. ComfyUI renders against a custom Bakery LoRA we trained. A Critic agent reviews against the style guide.

  • 1,112 assets · 105 events shipped in 2025
  • External art cost: €117,720/yr → on-platform with Symbiotica retainer
  • Brief → rendered pack: hours, not days
STILLFRONT·March 2026

8K Game Jam — AI asset workflows for 8 teams

8 teams across Stillfront's studios. 8 prototype games. ~36 hours. Razvan led the AI asset workflow — design briefs in, themed asset packs out, in time to actually use them in the jam build.

This was Symbiotica before it had a name. The validation run: can a small AI-assisted art pipeline keep up with 8 teams shipping concurrently? It could. The patterns from that weekend — the multi-step pipeline, the per-game LoRA, the agent-style coordination between brief and render — became Symbiotica's thesis.

  • 8 game prototypes shipped
  • Multi-step pipeline validated at studio scale
  • Direct lineage to Symbiotica, founded two weeks later
SYMBIOTICA (INTERNAL)·Ongoing

The platform itself

Imperia Hub. The LVL0 studio portal. The Symbiotica command center. The agent registry, the skill library, the runner. The Designer, Plan, Critic agents. All of it lives in our repos and ships through the same pipelines we sell to clients.

We use the platform on the platform. When we need a new pipeline step, the Plan agent triages the idea into a backlog card. When a new agent is useful, the meta-tools spin it up. When a skill needs refining, the Claude copilot drafts the next version. It's the surest test we know — if our own work doesn't depend on it daily, it's probably not worth selling.

  • Agents shipped: Plan · Designer · (more)
  • Self-improving by design
  • Same tools, same agents, same accountability