Latest research
How to Run an Optimizer Benchmark That Survives a Skeptic
Five parity rules that decide whether your speedup number is real, written from the buyer's side of the table.
Read post
Your Optimizer's Average Runtime Is the Least Interesting Thing About It
When a real deadline sits on the other side of the computation, the tail is the product and the mean is decoration.
Read postNobody Ships Machine Precision: Choosing the Accuracy Your Product Actually Needs
How to set an optimizer's accuracy target in basis points and dollars, and why grading it with an objective ratio quietly breaks.
Read post
Harvested Losses Are an Activity Number. Clients Keep a Different One.
Gross losses harvested measures how much you traded. What a named client actually keeps is a harder number, and it is now computable.
Read postHow to Prove a Performance Claim Without Open-Sourcing Your Engine
A contract for publishing proprietary performance claims that a buyer can check on the outputs, with tools the buyer controls.
Read postDeadline-Bounded Institutional Rebalancing: What It Is and Why It's Hard
A hard deadline is a constraint on the workflow, not a preference on the objective — and it changes what a good answer even means.
Read postInside the Asymmetry PRISM Benchmark: What We Measured and How
A plain-language walkthrough of the arXiv:2606.23367 methodology — the evaluation boundary, "completed rows", and what the number cannot tell you.
Read post
Optimizing Grid Dispatch Inside the Control Tick: The CAISO Case
Where the ~5 ms grid deadline comes from, and why feasibility and plan stability matter more than raw speed.
Read post
After-Tax Alpha: Why Tax Management Became the Battleground
When beta is free, after-tax outcome is the last reliable edge — and it rewards computation over intuition.
Read postWhat "Institutional-Grade" Actually Means for an Optimization Engine
The five testable properties that separate the real thing from the brochure — and the questions that reveal which you're looking at.
Read post
The Personalization Wave: Why Direct Indexing Is Eating the SMA
The market is shifting from pooled funds to personalized, tax-managed accounts faster than almost anyone is operationally ready for.
Read postThe Overnight Batch: Pricing a Whole Book Before the Open
Why nightly rebalance batches break at scale — and why deadlines are governed by tails, not averages.
Read post
A Practical Taxonomy of Portfolio Optimization Methods
Convex, integer, first-order, heuristic, quantum — a vendor-neutral guide to what each family is good at and where it breaks.
Read post
Wash Sales, Lots, and the Hidden Complexity of After-Tax Investing
Why tax-loss harvesting is a constrained optimization problem, not a spreadsheet exercise.
Read post
Build vs. Buy: The True Cost of an In-House Optimization Stack
The prototype is the cheap 20%. Scale, validation, audit, and years of maintenance are the rest.
Read post
Determinism, Reproducibility, and the Coming Audit Standard
Why reproducibility is becoming a compliance requirement — and what it takes to re-derive any trade on any date.
Read postReading Solver Benchmarks Like an Adversary
A buyer's checklist for tearing an optimization benchmark apart — and the questions the slide was built to skip.
Read post
GPU Computing in Quantitative Finance: A Sober Look
Where massively parallel hardware genuinely helps in finance — and where it quietly makes small problems slower.
Read postQuantum Optimization for Portfolios: Hype, Reality, and an Honest Roadmap
Where quantum methods might eventually help portfolio construction — and why honesty about the timeline is the credible posture.
Read post
Risk Models Meet the Optimizer: How Factor Structure Shapes Real Portfolios
How factor structure flows into construction, and why the risk model and the optimizer have to be co-designed.
Read postMore from the library
How to Read Portfolio Optimization Benchmarks Without Being Misled
Cold starts, repeated runs, quality gaps, and why matched workloads matter more than headline speedup.
Read postWhy Direct Indexing Needs Optimization Infrastructure, Not Just Screens
Personalized portfolios become an operations problem once taxes, lots, turnover, tracking error, and scale arrive together.
Read postWhat a Matched-Workload Pilot Should Prove
A practical checklist for evaluating any optimization engine against your current production workflow.
Read postTax-Loss Harvesting Latency Is a Product Constraint
Why harvesting quality depends on solving fast enough to evaluate lots, wash-sale risk, tracking error, and turnover together.
Read postWhen Portfolio Optimization Crosses the GPU Boundary
The practical case for GPU-native optimization once asset count, scenarios, and account batches stop fitting overnight workflows.
Read postWhat Quantum-Ready Means for Portfolio Construction
A grounded view of the research frontier for position limits, cleanup, and hard combinatorial portfolio edges.
Read post