Insights

The Specification

My Substack newsletter at the intersection of AI research and institutional capital allocation. Each issue translates frontier research, causal inference, simulation methods and model specification, into practical tools for CIOs, investment committees, and consultants evaluating AI-driven strategies. Infrequent but substantial, with frameworks you can use in your next manager meeting.

The State of AI Alpha

The industry's leading quarterly research report on AI in institutional investing. Each edition maps what works, what doesn't, and the emerging risks most allocators can't yet see, with evidence, not hype. Built for CIOs, manager selectors, and investment committees making real allocation decisions.

Below you will find a selection of my research contributions, commentary pieces, conference speeches, and media coverage.

Cheap Tokens, Costly Chips and a Missing AI Payoff

Bloomberg, Taking Stock

Bloomberg built its AI payoff column on my census of 919 earnings calls across the 60 largest US-listed financial firms. Four in five talked about AI, more than half discussed its cost, and exactly one firm ever named a realised dollar return. Also carried by the Los Angeles Times.

Inside Sovereign Wealth Funds' AI Push

AI Street

An interview on what AIMI found across 27 sovereign wealth funds and 26 pension funds: the largest asset owners are adopting AI quickly, and none of them can yet show it has made them money.

Would the Backtest Survive a Different Specification?

CFA Institute Enterprising Investor

Configuration risk: the same strategy, rebuilt with equally defensible settings, can produce a different track record. Published factors lose 26% of their return out of sample as an artifact of data mining, and more than 160 research teams handed the same data produced a spread of answers as wide as their standard errors.

The Question That Reveals Whether a Quant Manager Can Explain a Single Trade

CFA Institute Enterprising Investor

Explainability is not a nice-to-have for allocators, it is the difference between oversight and faith. What a manager should be able to say about one position, and what it means when they cannot.

Did the Manager Change the Model or Just the Settings?

CFA Institute Enterprising Investor

Calibration is not learning. Ask a manager to walk you through a time the model was wrong, then separate what they rebuilt from what they merely retuned. Twelve architectures can post the same Sharpe ratio with a threefold spread in turnover.

Same Sharpe, Different Beliefs: What Your Quant Manager's Optimizer Is Hiding

The Specification

MIT and Princeton research shows two managers with identical Sharpe ratios can hold opposite beliefs about how markets process information. The difference is invisible to standard due diligence, and it lives in the optimizer, not the variables.

71% of Active Management Is Pattern. Here's What the Other 29% Requires.

The Specification

New research from Harvard, ADIA Lab, and others reveals that a simple AI model can predict 71% of active fund manager trades in advance. The managers worth finding are the ones a neural net cannot explain, and their edge depends on simulation infrastructure that almost no one has built.

Votre gérant utilise l'IA. Vraiment?

Market.ch

Three questions that reveal whether a quant manager truly understands their AI or is riding the wave of AI washing. No engineering background required, just the right questions at the right moment.

The Question That Exposes Weak Quant Models

CFA Institute Enterprising Investor

Why specification error, not overfitting, may be the largest undiagnosed risk in systematic strategy evaluation. One question about variable inclusion and exclusion reveals whether a quant model is built on economic reasoning or statistical accident.

The SPEC Test: Four Questions That Reveal Whether a Quant Model Is Built on Evidence or Accident

A structured framework for testing whether the structure of a quantitative model is sound, not just whether its outputs have looked good historically.

AI as the Linchpin for Modern Asset Allocation

Beyond[Profit] Forum, Saudi Arabia

How artificial intelligence is reshaping portfolio management and asset allocation decisions for institutional investors. Presented at the Beyond[Profit] Forum in Riyadh.

Swiss Investors Should Pay Attention to the AI Compliance Technology Market

Bilan

Interview on AI regulation in Europe, the role of institutional investors in shaping AI policy, and why Swiss investors should pay attention to the AI compliance technology market.

The EU AI Act and Institutional Investors: What You Need to Know

Allnews

How the EU AI Act's risk classification framework affects institutional portfolios, from compliance costs to new investment opportunities in AI governance technology.

Representing Institutional Investors at the OECD Digital Trust Convention

OECD

Represented the perspectives of institutional investors on AI governance and digital trust at the OECD Digital Trust Convention, alongside UNESCO, IEEE, and Atlantic Council.

Explainable AI for Institutional Finance

Digital Finance MSCA Training Week

Why explainability is becoming a competitive requirement in institutional investing: the trade-off between model accuracy and interpretability, and what it means for allocators evaluating AI-driven strategies.