chini-013-pottery-studio

Pottery Studio Firing Schedule

Two kilns, twenty members, four firing stages, one electrical limit. Don't crack the work.

Source: Small ceramics studio operations, kiln scheduling literature, the author's friend who runs a studio in Brooklyn

Prompt

Design the firing pipeline for a member-run pottery studio.

Functional:
- Members drop greenware on the bisque shelf when ready. Each piece must move bisque -> glaze application (member work) -> glaze firing -> pickup.
- Two kilns: one bisque-only, one glaze-only. Each kiln cycle is 18 hours and runs only when full.
- Members can only glaze pieces that have been bisqued. The bisque shelf and glaze shelf are physically separate.
- Studio drops finished work on the pickup rack with member name.

Non-functional:
- A member-event week (4x normal output) must not cause kilns to skip safety underloading or members to walk out without their work.
- If one kiln fails, the other must NOT be reused for both stages (cross-contamination ruins glaze). The system rate-limits intake instead.
- The shared 200-amp panel cannot run both kilns at once. Schedule must enforce this.

Return a Chinilla CanvasState. Components are people, kilns, shelves, the schedule. Behaviors: queue (shelves), batch (kiln cycles), ratelimit (electrical cap), circuitbreaker (kiln failover refusal), split (bisque vs glaze routing).

Constraints

Max components: 12
Required behaviors: queue, batch, ratelimit
Monthly budget: $6000

Stress scenarios

Normal week

baseline

Steady member output, both kilns healthy.

Open studio event

spike

Member volume 4x for the week. Kilns must batch, shelf overflow must be prevented.

Bisque kiln fails

outage

Bisque kiln down. System must rate-limit intake, not reroute through glaze kiln.

Long cone-10 cycle

latency

Glaze firing extended for cone 10. Downstream shelf must absorb without dumping work.

Pass criteria (overall)

Min stability score: 60
Max drop rate: 10.0%
Min delivery rate: 85.0%
Max errors: 6

Submit your run

Submissions go through the chini-bench CLI. It calls your model with your key, scores the result locally, and posts to the leaderboard. Nothing leaves your machine except the canvas it produces.

End-to-end:

pip install git+https://github.com/collapseindex/chini-bench-cli.git
export OPENROUTER_API_KEY=...

chini-bench run chini-013-pottery-studio \
  --provider openrouter --model google/gemini-2.0-flash-001 \
  --as alice --x alice --linkedin alice-builds

Or inspect the prompt first:

chini-bench prompt chini-013-pottery-studio

Providers: openai · anthropic · google · openrouter · ollama

Leaderboard

Rank	Submitter	Model	Score	Stability	Design	Pass	Links
#1	alex default	O openai/gpt-5.4	47	61.0	75.0	✗	X
#2	alex default	G google/gemini-3.1-pro-preview	45	56.0	75.0	✗	X
#3	alex default	X x-ai/grok-4.20	29	21.0	75.0	✗	X
#4	alex default	A anthropic/claude-sonnet-4.6	24	9.0	75.0	✗	X

Per-scenario breakdown of the top run

Scenario	Health	Drop rate	Pass
baseline	53.0	31.3%	✗
member-event	50.0	37.2%	✗
kiln-down	86.0	2.8%	✗
slow-cycle	53.0	31.3%	✗

How is this scored? →