SimicX × Google × CBAIA — AI in the hunt for alpha
The half-life of alpha
is collapsing.
Same data. Same models. Same compute.
So where is the edge now?
Thanks everybody for participating. Follow us on LinkedIn for our next rounds in London and San Francisco.
For a decade a good signal lasted five to seven years. Today it decays in eighteen months — often far less.
The field has converged. The same data vendors, the same model families, the same clouds and the same GPUs. The moment you find an edge, the rest of the market is one fine-tune behind — and the edge you priced yesterday is consensus today.
So the question is no longer whether alpha decays. It is whether AI can find, generate and validate it faster than it dies. That is the whole evening — AI in the hunt for alpha, argued out by the people doing it.
Hours, not years
Signals that once paid for half a decade are arbitraged away in months — sometimes hours. Half-life is the only number that matters now.
t > 3.0
With hundreds of factors already mined, Harvey, Liu & Zhu argue a new one needs a t-stat above 3.0 — not the textbook 2.0 — before you believe it.
Most backtests are false
Bailey & López de Prado, bluntly: run enough trials and any strategy looks brilliant in-sample. Out-of-sample is where it goes to die.
≈42% more overlap
Manager books converged roughly 42% from 2013–2024 in 13F filings. Increasingly the same positions, held by everyone, at once.
AI is the accelerant
Train every desk on the same data with the same models and signal lifespans fall convexly with adoption. The tool that finds alpha also kills it.
Each crest is an edge found; each collapse, its decay. The wave only tightens — more alpha, found and lost, faster every year. Out-running it is the work.
LEAP
The leap of faith
you no longer have to take.
Backing a new edge has always been a leap of faith. The point of this room is to make it one you no longer have to take — find it, prove it, and measure its decay, before anyone commits a dollar.
Deliberately small. Senior. Off the record — people who live this problem and almost never get to compare notes.
SimicX, Google and CBAIA open the round. A panel of senior practitioners then takes on the questions you send us in advance about AI in discovering, generating and validating alpha, before the room turns to open networking and a roundtable.
Opens the round
AI in the hunt for alpha — what compounds, what decays, and what actually survives validation. A starting point, not a finish line.
Then · the panel
C-suite and senior portfolio managers from tier-one hedge funds and asset managers answer the questions guests submit in advance.
Then · the roundtable
The room opens into networking — guests form the groups they most want to talk in.
The panel
Francesco Filia
Founder & CEO
Fasanara Capital
Dr. Eric Huang
IAC Research Lead
Point72
Dr. Simon Wang
Founder & CEO
SimicX
Nathan Van Balen
Hedge Fund Lead
Google Cloud UK
ModeratorWho these seats are for
Portfolio managers, senior PMs, CIOs and CEOs at active funds — hedge funds, asset managers, family offices and proprietary trading desks — who are themselves systematic, quantitative practitioners, directly and materially engaged in money management or trading.
The bar: a live book of no less than US$100M under management. Peers, not vendors — the best at what they do.
Off the record
Held under the Chatham House Rule. Ideas leave the room; names do not. No recording, no slideware, no press — which is exactly why people speak plainly.
The essentials
Hosted with Google and CBAIA at 6 Pancras Square, by King's Cross St Pancras, London N1C 4AG. Wednesday 8 July 2026, 17:30–20:30, with food provided.
The room is shared with confirmed guests.
Thanks everybody for participating
LEAP 01 has concluded.
Please follow our LinkedIn page for updates, and we'll keep your name in mind for our next rounds in London and San Francisco.
Follow us on LinkedIn ↗LEAP 01 has concluded and registration is now closed. Please follow our LinkedIn page — we'll keep you posted on our next rounds in London and San Francisco.
Thanks everybody for participating
LEAP 01 has concluded and registration is now closed. Please follow our LinkedIn page — we'll keep you posted on our next rounds in London and San Francisco.
Follow us on LinkedIn ↗