The MorphIQ journal
Notes from the work.
Engineering decisions, numerical investigations, and the things we learn by building.
Open-Sourcing Tollgate: Bounded Spend Without a Network Hop per Request
Tollgate is an open-source Rust library for quota admission and usage accounting. Each instance spends a fenced lease locally, so a metered request is admitted, charged and billed without synchronous I/O, and the spend stays bounded by what the account was allocated.
The Boundary Is Part of the Algorithm
A numerical method needs a justified domain. The threshold that selects it deserves the same scrutiny as the formula itself.
How to Evaluate a Market Regime Detector Without Fooling Yourself
A detector can look fast simply because it changes its mind constantly. This held-out synthetic benchmark shows why regime research needs ground truth, causal features, strong baselines, and a scorecard that keeps latency beside stability.
How Spread Foundry Uses Regime Context
FerroWave turns trailing market history into a versioned evidence pack. Spread Foundry uses that context to explain suitability and risk without treating a regime label as a forecast or hiding structurally valid spreads.
Advanced Signal Decompositions Need the Right Test
SWT, synchrosqueezing, EMD, VMD, and EWT encode different assumptions. Four synthetic diagnostics show what each method earns, where a direct baseline still wins, and why one leaderboard cannot answer every regime question.
Deterministic Replay for Trading Systems
A trading system that cannot explain a run from its recorded inputs is hard to debug and harder to trust. Deterministic replay turns reproducibility into a runtime contract.
Options Spread Analytics Should Be a Workflow, Not a Table
Spread analytics becomes useful when it moves from isolated metrics to a guided workflow: market posture, symbol context, strategy fit, candidate ranking, and risk review.
Price-Time Priority Matching in Rust
A matching engine is judged by latency, but the real contract is deterministic priority: price first, time second, with explicit cancellation, tie-breaking, and replay semantics.
Rust Market Data Normalization Starts at the Boundary
A market-data normalizer should not be a pile of exchange adapters. The useful abstraction is a canonical event boundary with explicit time, precision, quality, and replay semantics.
What a Rust Options Pricing Engine Has to Prove
A fast options-pricing engine is not just a collection of Greeks. The contracts around monotonicity, parity, no-arbitrage bounds, implied volatility, and rounding behavior matter just as much as throughput.
Wavelet Feature Extraction for Time-Series ML Pipelines
Wavelet transforms are useful before prediction begins: they turn non-stationary time series into scale-aware features that can feed ML and AI pipelines without hiding the time structure.
Determinism as a Correctness Contract
Why Anvil treats bit-for-bit reproducibility as a testable property — and how an event-sourced core makes a backtest and a production run literally the same program.
SVI and the No-Arbitrage Volatility Surface
The raw SVI parametrization is five numbers per slice. Keeping a fitted surface free of butterfly and calendar arbitrage is where the real work — and the real math — lives.
Wavelet-Domain Regime Detection with the MODWT
Markets change character at different speeds. The MODWT splits a return series into scale components whose variances add up exactly — which turns a vague question about regimes into an estimable one.