Investor memo · Sep 2026 · Confidential Contact
Abstract FinanceThe hedge fund, rebuilt as software

The hedge fund, democratized.

The best investment strategies, available to everyone. And the next great fund manager may be an AI.

Abstract is where human and AI investors build public track records, compete for capital, and manage money for anyone.

Anyone can build an investor. The best ones get funded.
Live manager · illustrative
Semiconductor
Intelligence
Autonomous AI · live 14 months · 3,421 investors
Autonomous
+27.8%S&P 500 +16.1% · +11.7 pts
Max DD
−9.4%
Vol
13.1%
AUM
$38.2M
Fee
1.25%
$10,000Allocate
Every trade on the record
The next great investors won't start hedge funds. They'll launch on Abstract.
Hedge fund capital · HFR
$5.6T
Global AuM · BCG
$147T
Minimum to follow a manager
Your choice
Client assets we hold
Zero
Stage
Building · pre-launch
01The democratization

For fifty years, the best strategies were private. Abstract opens the menu.

Accreditation rules, high minimums, subscription documents and quarterly letters. The most sophisticated active management in the world has been built for institutions and the wealthy.

Before · family office, $50M
"Give me your best semiconductor manager."
  • Sector specialists and long/short funds
  • Institutional research processes
  • Thousands of differentiated managers
Before · retail investor, $25K
"Here's an ETF."
  • An index fund
  • A robo-adviser
  • Picking stocks alone, or copying a personality
After Abstract:
the same menu.

The same managers, the same live records, the same rules, whether you invest $1,000 or $100 million.

Online brokers
Opened access to stocks.
ETFs
Opened access to diversified portfolios.
Abstract
Opens access to active management.
02Why now

AI is creating a new species of investment manager.

Building a fund used to mean a PM, a team of analysts, data, engineers and operations. An AI-native manager needs a frontier model, the right data and a clear investment philosophy.

Traditional hedge fund

  • Portfolio managerHuman
  • Team of analystsHuman
  • Data teamHuman
  • EngineersHuman
  • Operations, complianceHuman
Years to build · millions to run

AI-native manager

  • Frontier modelCompute
  • Proprietary dataData
  • Investment philosophyJudgement
  • Research, around the clockCompute
  • Operations, complianceAbstract
Days to build · compute to run

The next great investment manager may be an AI.

Yesterday

Investment talent was a small professional class of fund managers.

Today

AI makes it possible for millions of different investment strategies to exist.

The gap

There is no trusted way to find out which of them actually have skill.

Abstract

Let them compete. Measure everything. Capital follows the winners.

03The problem

Finding investment skill is now harder than creating it.

AI lets anyone manufacture a brilliant-looking backtest. Run 10,000 variations, publish the winner. The only evidence that still means something is a live record with real money, kept by someone else.

Backtests are cheap.

Infinite, free, and easy to overfit. On Abstract they are shown separately and never blended into a ranking.

Live capital is proof.

Every manager starts with their own money. Every trade is recorded by Abstract, including the versions that failed.

01

Sandbox

Unlimited simulation and backtesting.

No claims shown
02

Live

The manager trades it with their own money.

Own capital at risk
03

Provisional

Public after a first live week, clearly flagged.

Short-history label
04

Qualified

Enough live history to take investor capital.

Open to investors
05

Established

Proven AUM, history and adoption. More visibility, more capacity.

Featured
Ranked on
Risk-adjusted returnMax drawdownConsistencyAcross regimesLive vs backtestStrategy ageCapacityTurnoverHuman interventionAUM and investors
04Abstract

The arena for investment intelligence.

Human analysts. Quants. Domain experts. Autonomous AI. Everyone competes under common rules, and the winners earn capital.

01

Own money first

Every manager starts with their own capital at risk.

02

Every trade recorded

Independently, by Abstract, not reported by the manager.

03

Common rules

Same mandate limits, same metrics, same benchmark logic.

04

Capital follows

Investors allocate to the managers that earned it.

LiveLeaderboard · track records
AllAutonomousSystematicHumanHybrid
#ManagerType · stageLiveReturnvs S&P 500Max DDAUMTrend
Ranked on live, risk-adjusted results. Backtests shown separately, never blended in.Illustrative data
05For investors

The same managers. Whether you invest $1,000 or $100 million.

Choose a manager, choose an amount, one click. Your money stays in your own brokerage account and follows the manager's portfolio automatically.

×No accreditationOpen to any investor
×No high minimumFractional shares
×No subscription docsAn account, not a fund
×No lock-upExit any trading day
×No poolingYour assets, your account
×No black boxEvery trade and decision logged
STEP 01

Browse

Live records, risk, benchmark, strategy.

STEP 02

Choose

A manager and an amount.

STEP 03

Follow

Your account mirrors the target portfolio.

ANY TIME

Leave

Reduce or exit whenever markets are open.

↳The architecture

One manager. One decision. Any number of investors.

A manager publishes one target portfolio. The brokerage layer fans it out across every account that follows. AI compute scales with the number of managers, not with assets or investors.

Target portfolio · signed
NVDA10.0%
TSM10.0%
ASML8.0%
GOOG5.0%
…
CASH15.0%
06For managers

Describe how you invest. Abstract builds the manager.

Anyone or anything that can produce alpha can become an asset manager. Plain language in, an executable strategy out, or bring your own agent. Abstract tests it hard, then it earns a live record.

Quant researchersEx-hedge-fund analystsIndustry executivesProfessorsDevelopersFinance writersProfessional PMsAutonomous AIResearch agents
Strategy brief
+ Research PDFs+ Data feeds+ Alt data+ Prompts+ Python+ Your own agent+ Any model
Automatic evaluation
Historical backtest
Walk-forward test
Unseen periods
Regime analysis
Factor attribution
Drawdown analysis
Costs and slippage
Turnover and capacity
Parameter sensitivity
Perturbation tests
Look-ahead bias check
Overfitting diagnostics
Go live with your own capital. The verified record starts now.Live

Managers earn on the capital that follows them

Assumption · 0.60% manager share
07The launch

Abstract Arena. Which AI can actually invest?

40 AI investment managers.$200,000 of real capital.One public leaderboard.

20 built by Abstract, 20 by outside builders, $5,000 each, invested rather than spent. Visitors can follow a manager, or say they want to invest before we take any outside money.

FollowI want to invest
No cherry-picked backtests
No deleting failures
No changing history
Every decision recorded
40→400→4,000→40,000 managers

Kaggle's competition, Y Combinator's funnel and a hedge fund seeding platform, with performance measured in real dollars.

08AgentOS

Intelligence can be creative. Execution is boring on purpose.

Abstract runs the agents: models, memory, data, triggers, backtests, monitoring, failover. Whatever happens inside, a manager can only say one thing: this is the portfolio I want.

Filings, transcripts, market and own dataInputs
↓
Research layerClaude, LLMs, tools
Persistent strategy stateTheses, memory
Decision layerModels or manager logic
↓
Signed target portfolioThe only output
↓
Deterministic control layerMandate, kill switch
↓
Manager of record → broker → investorRegulated partners

The AI can be wrong. It can never be dangerous.

A hallucination can cause a bad trade inside the mandate. It cannot move client money or leave the rules.

Long-only equitiesNo leverageNo optionsNo penny stocksMax 10–15% per nameLiquidity floorsTurnover capsPrice sanity checksStale-signal expiryNo transfers, everImmutable audit log
No decision = hold.

If a model, data feed or Abstract itself goes down, positions stay put and investors can still exit at the broker.

Never custody.

If Abstract disappeared tomorrow, every investor would still own their stocks at the broker.

# Bring your own agent: three endpoints
GET /health
GET /version
GET /target-portfolio
09Business model

Paid like an asset manager. Scales like software.

Investors pay a percentage of assets, split between the manager, Abstract and the regulated partners. The 100,000th investor following a manager costs almost nothing extra to serve.

0.60%
0.40%
0.25%
$300

Manager. Licenses the model. Upside grows with capital.

$200

Abstract. AgentOS, arena, records.

$125

Partners. Manager of record, broker, custody.

$50,000 following one manager at a 1.25% fee = $625 a year. Illustrative split, set with partners

Cost follows managers. Revenue follows assets.

Schematic, not to scale.

Managers bring their own capital

  • Finance writers and their readersAudience
  • Ex-hedge-fund analysts and PMsNetwork
  • Industry executives, academicsCredibility
  • Quant and AI developersSupply

A manager's own network brings the first capital. The leaderboard brings the rest once the record is proven.

Better managers→better returns→more capital→bigger payouts→Repeat
10Regulated by design

Not regulatory arbitrage. A clean separation of roles.

Managers and Abstract supply software and models. A registered manager of record holds the client relationship and the discretion, and supervises the mandate, not every investment thesis. That is what lets manager #10,001 launch as software.

Model provider

Manager

  • Licenses a generic model
  • No client data
  • No access to assets
Software

Abstract

  • AgentOS
  • Arena and leaderboard
  • Verified records
Manager of record

RIA (US) · MiFID manager (EU)

  • Discretion
  • Mandate limits
  • Kill switch
Infrastructure

Broker and custodian

  • KYC, funding
  • Execution
  • Custody, tax docs
Owner

Investor account

  • Individual account
  • Own holdings
  • Exit any trading day

Every piece already works somewhere. Abstract puts them together.

Model licensing · US

dub

Model providers license strategies to an SEC-registered adviser, which keeps full discretion over client accounts.

Agentic trading · May 2026

Robinhood

Third-party AI agents trade dedicated retail accounts without approving each order.

Model marketplaces · US

Vestmark · SMArtX · Orion

Thousands of third-party models implemented across client accounts by a regulated overlay manager.

Copy-trading guidance · EU

ESMA

Automatically copying a signal provider is portfolio management, so a MiFID manager sits in between.

Partner pipeline

Outreach under way · Sep 2026
United States
AtomicVestmarkSMArtXAdhesionGeoWealthOrionAssetMarkEnvestnetBNY Pershing · Wove
Europe
Atomic Europelemon.marketsGinmoninvestify TECHUpvestWealthKernelFNZ
The one question for every partner: can manager #10,001 launch without becoming a new regulatory project?

If no partner can say yes, Abstract registers its own RIA. Planning estimate: 2–4 months.

11The long game

Abstract discovers investment intelligence.

Five years in, Abstract holds complete live histories for tens of thousands of managers. That is a dataset no allocator has ever had.

Which model architectures work
Which data creates alpha
Which humans have skill
Which agents survive regime changes
Which styles collapse together
How much capacity each one has
Funds of agents

The 30 best uncorrelated managers on Abstract, in one allocation.

Abstract stops being only a way into existing strategies. It becomes a better mechanism for allocating capital to investment intelligence than the hedge fund industry has had.

What compounds

Track records

Years of verified live history cannot be recreated overnight.

Managers

The best ones have their AUM and income on Abstract.

Capital

A manager with $200M following it does not switch platforms.

The dataset

What works out of sample, with real money, across humans and machines.

AgentOS

The easiest place to build an autonomous investment manager.

Regulatory OS

A standard path from software strategy to retail managed account.

Phase 1

Arena

40 live AI managers, public records, demand signal.

Phase 2

Regulated launch

Manager-of-record partner, investor capital in the US and EU.

Phase 3

Open creation

No-code building for anyone, bring-your-own-agent for pros.

Phase 4

Funds of agents

Allocation products built on the dataset. Own licences where they pay off.

12Team

Bridgewater. Then built and sold a venture-backed company.

Markus
Maier
Founder
LinkedIn ↗

Already built: a production systematic trading engine with broker execution, risk parity, regime detection, hedging and hard kill switches, paper trading since December 2025.

  • FinanceBridgewater AssociatesThe world's largest hedge fund
  • Built & soldNudge, sold 2026Backed by Balderton, Coinbase, Brevan Howard
  • OperatedAxel Springer SE, Chief of Staff
  • ProductAmazon Prime Video, Senior Product Manager
  • TrainedHarvard Business School · HV Capital
13The ask

Raising our first round to launch the Arena, build AgentOS and sign the regulated partner.

Arena capital

Real money behind the first 40 managers.

AgentOS

Building, hosting, testing and the control layer.

Regulated launch

Manager-of-record partner, counsel, compliance.

First hires

Engineering and quant, lean and senior.

markus@abstract.finance →

The next great investors won't start hedge funds. They'll launch on Abstract.

The hedge fund, democratized. Benchmarked, transparent, open to everyone.

AAppendix

Diligence.

Assumptions in this memo
Manager page, leaderboardIllustrative data, not live resultsIllustrative
1.25% fee and 0.60 / 0.40 / 0.25 splitWorking model; final split depends on partners and regulationAssumption
Manager income tableAUM × 0.60%, before costsAssumption
$5,000 own capital to go liveStarting point; smaller amounts possible for experimentsAssumption
Arena: 40 managers × $5,000Plan, 20 internal and 20 externalPlan
40 → 40,000 managers; funds of agentsAmbition, not a forecastVision
Own RIA: 2–4 months, ~$30–75k setup, ~$40–100k+ a yearPlanning estimates, not fee quotesEstimate
Managers bring early capital from their networksHypothesis the Arena testsHypothesis
What we still need to prove
  • Supply: non-developers can turn investment knowledge into working agents; professionals trust Abstract with proprietary logic.
  • Quality: agents behave reliably with real capital; we can separate skill from luck and overfitting.
  • Demand: investors allocate meaningful amounts to verified managers.
  • Regulation: a manager of record approves the framework rather than each strategy; managers can legally receive AUM-linked licensing economics.
  • Economics: Abstract keeps enough of the fee after managers, partners, compliance and support.
Why not crypto rails
Public-chain execution would publish every rebalance in real time, so good strategies could be copied or front-run without paying the fee. It adds smart-contract, key and upgrade risk, and tokenising a security does not remove securities regulation; ESMA applies its copy-trading logic to crypto services too. Crypto may later be useful as settlement infrastructure. It is not the core thesis.
Sources
  • HFR, Global Hedge Fund Industry Report, Q2 2026: industry capital $5.6T.
  • BCG, Global Asset Management Report 2026: global AuM $147T in 2025.
  • Robinhood, Agentic Trading launch, May 2026, and support documentation.
  • dub Advisors, Premium Creator Program and Form ADV Part 2A.
  • ESMA, supervisory guidance on copy trading under MiFID II.