jane-street_skill

Overview

This skill teaches how to design and implement trading systems in the style of Jane Street: functional, strongly-typed, and correctness-first. It emphasizes making illegal states unrepresentable, explicit effects, immutability by default, and rigorous testing for mission-critical market-making and pricing engines. Use the guidance to build safer real-time components, pricing pipelines, and risk checks even when using Python as the implementation language.

How this skill works

The skill inspects system boundaries, state machines, error surfaces, and side-effect locations, and translates Jane Street design patterns into practical engineering rules. It guides you to encode states as explicit variants, use result-style error propagation instead of exceptions, prefer pure functions for pricing, and isolate mutable state for real-time pipelines. It also prescribes property-based testing, incremental computation patterns, and modular abstractions to improve reasoning and correctness.

When to use it

• Building market-making systems, pricing engines, or execution infrastructure.
• Implementing risk checks, order lifecycle handling, or exchange adapters.
• Designing real-time pipelines where incremental recomputation matters.
• When you need high-assurance code with minimal runtime surprises.
• Refactoring legacy trading code to reduce bugs and implicit state.

Best practices

• Model all domain states explicitly (sum types / enums) so invalid states cannot occur.
• Keep pure functions for calculations and side-effect boundaries small and explicit.
• Return result-like values for recoverable errors; avoid using exceptions for control flow.
• Favor immutability; contain mutable state and make updates explicit and local.
• Use property-based tests for invariants and QuickCheck-style generators for edge cases.
• Design composable modules with clear interfaces and exhaustive pattern matching.

Example use cases

• Design an order lifecycle where Pending, Acked, PartiallyFilled, Filled, Cancelled, and Rejected are distinct variants with safe transitions.
• Implement a Black–Scholes pricing module as pure functions with deterministic greeks and an optional implied-vol routine.
• Build a real-time pricing graph that recomputes only affected nodes when market data changes (incremental approach).
• Create risk checks that return structured errors (insufficient buying power, position limit exceeded) and force callers to handle each case.
• Replace ad hoc nulls and sentinel values with explicit option/result types for clearer invariants and safer composition.