AI-Native · Systematic · London
Human judgment is the
most expensive mistake
in financial markets.
AvB Capital is an AI-native systematic manager based in London. We invest across public equities using a research engine that identifies, tests, and deploys signals — without a human investment decision in the chain.
The Problem
89.5% of active managers underperform over 15 years. Not a bad run — a structural failure.
Zero of 22 equity categories beat their benchmark net of fees. (SPIVA Year-End 2024.) Their selling decisions are worse than random — 783 institutional portfolios: buying shows skill, selling loses 50bps per month versus random selection. No learning effect. Failure is stable across all experience levels. (Akbas et al., Journal of Finance 2023.) $60 trillion is still managed this way.
Active managers underperform over 15 years
SPIVA 2024
Still in active management — running on human judgment
Of trades now algorithmic — research layer hasn't caught up
"It's not that the wrong people are running money. Human cognition — under pressure, at speed, with incomplete information — is structurally outclassed. You can't hire your way out of it."
The Response
90% of trades are already algorithmic. We closed the last gap — the research layer.
The industry automated execution years ago. What it didn't automate — because it's harder, and because it threatens the people at the top — is research and alpha generation. That's the gap AvB was built to fill.
We search for the gap between what is knowable about a company — across its numbers, its language, and its fundamental trajectory — and what the market has already priced in. Every edge we trade has been discovered, stress-tested, and validated before it touches capital.
Every signal passes a seven-test statistical framework. 81% of what we research is rejected. What survives is sized and deployed inside a deterministic risk framework — no discretionary step between signal and trade.
The Machine
A machine that reads everything, reasons across all of it, and has no ego.
The edge is not in any one signal. It is in holistic judgment across all three inputs — at scale, continuously, without cognitive fatigue or career risk.
Numerical
Price · Factor · Structure
30+ years of survivorship-free pricing. 70M+ rows. Factor signals without the heuristics that bias human judgment.
Textual
Language · Tone · Narrative
850K+ LLM-extracted rows from earnings calls, 10-Ks, 8-K filings. Tone, promises, narrative change, moat — processed at scale, continuously.
Fundamental
Accounts · Moat · Quality
Point-in-time correct financials. Forensic accounting signals. The numbers — without the analyst's attachment to names they have already pitched.
Where We Come From
Innovators at the intersection of AI and financial services.
The team behind AvB Capital built and deployed AI systems into some of the world's largest asset managers before turning the same tools inward. NLP at institutional scale. Quantitative signal research across decades of market data. The same rigour — now applied to running the fund.
Capital markets
Years spent in equities in the City — understanding how institutional money moves, where the inefficiencies live, and why human judgement is the bottleneck.
Enterprise AI
Built NLP infrastructure deployed into major asset managers to process investment research at scale — before applying that same capability to signal generation.
AvB Capital
The convergence: a fund designed from the ground up for AI-native operation. No legacy structure to work around. No human layer to protect.
Why Ant vs Bear
The ant is systematic, tireless, and compounds small advantages into large outcomes over time. The bear relies on instinct and force. We are building the ant's approach to investing — rigorous, repeatable, and structurally better than the alternative.
What we are not
We are not a large institution. We are not a quant fund that added an AI layer. We are not trying to replicate what Citadel or Two Sigma built — they built brilliant human organisations. We built a different thing entirely.
What we are building
A systematic manager that gets materially better every week — because every trade result, every signal decay curve, every regime shift feeds back into the research engine. The compounding is structural, not incidental.