Engine Cells

Four working engine cells. Each wraps an external system and exposes its state as Quilt cells. Paper trading with real BTC prices. Backtest of 3 alternative universes. A grid-world (Minecraft in miniature). A Jupyter notebook. A knowledge base.

📈 Paper Trading

Real BTC price from CoinGecko. A simulated portfolio. A formula cell computes P&L. The cell updates as the price moves.
trading.btc_price · number (live)
trading.position · number (formula)
trading.pnl · number (formula)
—
source: coingecko
updated: —
uptime: —
position
0.5 BTC
bought @ $40,000
P&L (USD)
—
—

🔁 Backtest · 3 Universes

Same BTC price, three different strategies. Each universe is a Quilt cell. The chart shows all three equity curves.
universe.base · buy on MA(20) > MA(50)
universe.alt1 · base + RSI filter
universe.alt2 · exit on +5% / -2%
universe.base
—
universe.alt1
—
universe.alt2
—

⛏️ Grid-World (Minecraft in miniature)

A 12×12 grid. Cells are tiles. You are a player. Move with arrow keys or buttons. A listener cell fires when you step on a special tile.
world.player · position
world.tile.{x,y} · tile state
world.events · event log
grass
stone
water
fire
treasure
you

📓 Jupyter Notebook

A live notebook kernel wrapped as cells. Click "Run cell" to execute. The output becomes a cell. The state persists across cells.
jupyter.kernelspec · python3
jupyter.exec_count · number
jupyter.last_output · string
In [1]
name = "Quilt"
version = "0.3.0"
cells = 13
Out [1]
(executed)
x = 13 * 7
Out [2]
awaiting execution
print(f"{{name}} v{{version}} has {{cells}} cells and {{x}} is the answer")
Out [3]
awaiting execution
exec count: 1
kernel: python3

📚 Knowledge Base (NotebookLM-style)

Notes wrapped as cells. Ask a question — semantic search retrieves the most relevant notes, and an LLM generates an answer with citations.
kb.note.{i} · string
kb.embed.{i} · vector
kb.answer · string (LLM)
Ask: —
—

🔗 The cell graph (engine cells)

All 5 engine cells wired together. The trading price feeds the backtest. The backtest runs strategies. The Jupyter cell analyzes the results. The KB answers questions. The cell model is the substrate.
# All 5 engines as cells, in one sheet
- id: trading.btc_price
  kind: engine
  engine: coingecko

- id: universe.base.equity
  kind: engine
  engine: backtest
  strategy: ma_crossover
  data: trading.btc_price

- id: universe.alt1.equity
  kind: engine
  engine: backtest
  strategy: ma_crossover_rsi
  data: trading.btc_price

- id: universe.alt2.equity
  kind: engine
  engine: backtest
  strategy: ma_crossover_fixed_sl_tp
  data: trading.btc_price

- id: world.player
  kind: engine
  engine: grid_world

- id: jupyter.last_output
  kind: engine
  engine: jupyter

- id: multiverse.summary
  kind: formula
  value: '`Best universe: ${max(universe.base, universe.alt1, universe.alt2)}`'