Skip to content

What we build

Investment agent

It works to each portfolio’s charter: it evaluates opportunities, selects them and manages them inside its limits, every judgement sealed before its outcome.

Research and development phase

A thesis card · An illustrative company · within execution limits

Illustrative demo
Charter
The clauses that admit it: strategy, universe, limits
Base
The base rate of a comparable cohortvalue withheldname withheld from the agent that built it
Evidence
  • Quarterly disclosure [p. 14, table 2]· raised the beliefvalue withheld
  • Liabilities note [p. 22, note 9]· lowered itvalue withheld
Weekly cap: not reached
Counter-search
DelayFinancingLitigationCompetitionExecutionCustomersNothing found, recorded
Opportunity
  • Quality of the basisvalue withheld
  • Probability of the thesisvalue withheld
  • Opportunity at today’s pricevalue withheld
Risk
Its share of the portfolio’s riskvalue withheldinside the budget, by the risk agent
Execution limits
Sizevalue withheldunder the execution ceilingPolicy gate: passed
Exit conditions
The announced launch is postponedFinancing on worse terms than announcedWritten before entry; probed by code before any news
Sealed record
Probability sealed before the outcomevalue withheldand scored after it

A person can stop it at any moment

Illustrative: the company and the references are an example; values withheld.

For every investment it keeps a thesis it can be held to: where it started, what moved it, and when it ends.

What it keeps

  • A thesis for every investment: its base, its evidence, its exit conditions
  • Every execution inside its portfolio’s charter, on record
  • A record of sealed judgements, scored after each outcome

How it manages an investment

A charter the institution writes, then one company through the six steps, in the application as it will work.

The institution sets

The charter

The institution writes the portfolio’s objective, benchmark, strategy and limits; the agent cannot change them.

The charter is the investment policy statement written as rules: the universe it selects from, the risk budget, the authority. Code checks it before every order, and every amendment is a new signed and dated line.

QuantaupPortfolio charter
Illustrative demo
Equity portfolio 01Approved by the investment committee
Objective
Beat the benchmark over three years
Benchmark
An illustrative global equity index
Strategy
Quality companies at a reasonable price, on dated theses
Universe
Large listed companies with deep liquidity
Risk budget
Volatility 12% a year · drawdown 15%
Limits
One position 8% · one sector 20% · the agent’s share 40%
Under stress
Ceilings halve, and no new position opens
Stop
Any committee member, at any moment
Authority
  • AloneTrims, exits, raises inside the ceilings
  • With approvalA new position
  • EscalatedAnything outside the charter

Signed and dated · every amendment a new line

Illustrative demo only: simulated data, not a view on any company named or a recommendation, and Quantaup has no relationship with the companies named.

Evaluates

  1. Coverage

    It reads everything published about the portfolio’s universe as it appears, and links it to every position it touches.

    Every document by its publication date, every figure with its location, and every item routed to the positions it touches: directly, or through a supplier, a customer or a competitor.

  2. The thesis

    It starts from the rate of comparable companies, and each piece of evidence moves it by its weight.

    The base is the rate of a cohort of comparable companies, and the agent that builds it is not told which company it is. Evidence moves it through one tested function: a weight times the log of a likelihood ratio, damped by how solid the base is. A move past the weekly cap goes to a human review, so it never overreacts to the latest item. Then a separate agent searches for what would break the thesis, in classes declared beforehand, and “nothing found” is recorded as a result.

QuantaupSimulated executionCoverage
Illustrative demo

Universe: the portfolio’s charter

  1. 09:30T1Quarterly filingDirectSiemensMoved upThesis updated
  2. 09:34T1Central bank decisionWhole bookNo changeExits checked: far
  3. 09:39T2Sector data releaseSectorSiemens, Schneider ElectricMoved downThesis updated
  4. 09:43T1A supplier’s quarterly filingSupplierSiemensNo changeRecorded
  5. 09:48T1A competitor’s resultsCompetitorSchneider Electric, SiemensNo changeRecorded
  6. 09:52T2Industry surveySectorSiemens, Schneider ElectricMoved upThesis updated
Illustrative demo only: simulated data, not a view on any company named or a recommendation, and Quantaup has no relationship with the companies named.
OPP-0142Siemens
Illustrative demo

Thesis: demand passes its written level by its date.

T3Single-source report, unconfirmed

Weight after the brake +0.06·Belief from 30% to 31%

20%25%30%35%What the price assumes 25%Base123456
Belief
31%
What the price assumes
25%
The gap
+6 pts
  1. The base: a comparable cohort’s rate25%
  2. Quarterly filing[p. 14 (example)]+0.28
  3. Sector data release[sector release (example)]-0.20
  4. Industry survey[survey (example)]+0.18
  5. Press commentary[opinion, no primary source]0.00
  6. Counter-search: nothing found[five declared classes]0.00
  7. Single-source report, unconfirmed[one source (example)]+0.06
Illustrative demo only: simulated data, not a view on any company named or a recommendation, and Quantaup has no relationship with the companies named.

Selects

  1. The opportunity

    The opportunity is the gap between its belief and what the price assumes.

    A strong basis does not make a thesis likely, and a likely thesis is not an opportunity at any price: three quantities kept apart. What the price assumes is read from the expectations built into it, by a reverse valuation or from option prices.

  2. Construction and risk

    It sizes each position by its opportunity and its risk, inside the portfolio’s risk budget.

    It works out the position’s effect on the portfolio’s volatility, concentration and correlations, and tests the portfolio under stress scenarios. The risk agent is not the one that selected: it can cut, never raise.

Simulated executionScreening
Illustrative demo

Belief against price

OPP-0142 SiemensThe agent’s belief 31% · What the price assumes 25% · +6 pts

20%20%40%40%60%60%No gapThe agent’s beliefWhat the price assumes
Universe: the portfolio’s charter · Threshold: 5 pts
CandidateCriteriaVerdict
Enough base: PassedCounter-search: PassedGap over threshold: PassedLiquidity within execution limits: PassedRoom under sector ceiling: PassedRisk within budget: PassedEnters
Enough base: PassedCounter-search: PassedGap over threshold: FailedLiquidity within execution limits: PassedRoom under sector ceiling: PassedRisk within budget: PassedSet aside
Enough base: PassedCounter-search: FailedGap over threshold: PassedLiquidity within execution limits: PassedRoom under sector ceiling: PassedRisk within budget: PassedWaits
Enough base: FailedCounter-search: PassedGap over threshold: FailedLiquidity within execution limits: PassedRoom under sector ceiling: PassedRisk within budget: PassedSet aside
Enough base: PassedCounter-search: PassedGap over threshold: PassedLiquidity within execution limits: PassedRoom under sector ceiling: PassedRisk within budget: PassedEnters
Enough base: PassedCounter-search: PassedGap over threshold: FailedLiquidity within execution limits: PassedRoom under sector ceiling: PassedRisk within budget: PassedSet aside
Enough base: PassedCounter-search: PassedGap over threshold: FailedLiquidity within execution limits: PassedRoom under sector ceiling: PassedRisk within budget: PassedSet aside
Enough base: PassedCounter-search: PassedGap over threshold: FailedLiquidity within execution limits: PassedRoom under sector ceiling: PassedRisk within budget: PassedSet aside

OPP-0142 SiemensEnters

The agent’s belief31%
What the price assumes25%

+6 ptsIt passed every criterion, and only the gap is sized.

Illustrative demo only: simulated data, not a view on any company named or a recommendation, and Quantaup has no relationship with the companies named.
Illustrative demo

Volatility, against the charter’s budget9.1% budget 12%

Largest shares of the portfolio’s risk

  • Siemens24%
  • Schneider Electric19%
  • Microsoft17%

Siemens and Schneider Electric: one sector, counted together

Risk agent · apart from selection · cuts, never raises

Illustrative demo only: simulated data, not a view on any company named or a recommendation, and Quantaup has no relationship with the companies named.

The risk budget

A separate agent measures each position against the whole portfolio; it cuts, never raises.

Illustrative demo

Distance to the drawdown budget

  • Rates rise sharplyInside
  • Its sector sells offNear
  • Liquidity dries upInside

Stress scenarios

The portfolio is tested before every order, its distance to the budget in view.

Manages

  1. Execution

    Every order passes the policy gate, then an algorithm works it and its cost is measured.

    No model sends an order to the market: models propose a target portfolio, and code checks and executes it. The ceilings are in code, and configuration can only lower them.

  2. Monitoring and account

    It checks its exit conditions before every news item, and every judgement is scored after its outcome.

    The exit conditions were written before entry, and code checks them before any news is read. Probabilities are sealed before their outcomes and scored after them, every result is split into market, factors and judgement, and every failure comes back as a training environment.

Simulated executionPortfolio
Illustrative demo
Exposure
34.5%of a 40% share
Positions
6
Largest position
8%ceiling 8%
Uninvested in its share
5.5%

Daily exposure inside the agent’s share

Today·Exposure 34.5%

10%20%30%40%The agent’s share 40%29 days agoToday

Limits

  • One-position ceiling8% / 8%
  • Sector ceiling12.5% / 20%
  • The agent’s share34.5% / 40%

The ceilings are in code; configuration can only lower them.

Positions, and the one-position ceiling

  • SiemensIndustrials8%
  • MicrosoftTechnology6.5%
  • IberdrolaUtilities6%
  • WalmartRetail5.5%
  • Schneider ElectricIndustrials4.5%
  • PrologisReal estate4%

One-position ceiling 8%

Illustrative demo only: simulated data, not a view on any company named or a recommendation, and Quantaup has no relationship with the companies named.
Illustrative demo

SiemensDistance to the condition

  • Demand has not passed its written level by its dateFar
  • The margin falls below its written floorFar
  • The belief falls below what the price assumesNear

Written before entry, and checked by code before any news is read.

Illustrative demo only: simulated data, not a view on any company named or a recommendation, and Quantaup has no relationship with the companies named.

Exit conditions, before entry

Every position has its conditions in writing, and the distance to each in view.

Illustrative demo

Running inside the execution limits

Try it: every window on this page stops.

Stop at any moment

One button stops the whole agent, and the record keeps who stopped it, and when.

Illustrative demo

A resultDrawn without values

  • Market
  • Factors
  • Judgement

Judgement

  • Thesis
  • Timing
  • Size
  • Execution
  • Every sealed probability scored after its outcome
  • Every failure comes back as a training environment

The account

Every result is split, so the judgement’s part is known.

Move the evidence. Watch the judgement.

Evidence moves the belief, the cap brakes it, the price decides the opportunity, risk sizes it, and the execution limits cut the size.

An illustrative company

Illustrative demo
The base

How solid the base is

This week’s evidence
The counter-search
The price
Risk

The stock’s volatility25% a year, an example

Execution

One position’s ceiling: 8.0%

Exit conditions, written before entry

Belief in the thesis

35%was 25%

Why it moved

  • The quarterly filing: demand above what the thesis assumed+0.28
  • A large customer announces the contract on its side+0.18

Size

By the opportunity alone3.8%
After its risk2.3%
After the execution limits2.3%
  1. The base: the cohort’s rate is 25%, and its solidity keeps 40% of any move.
  2. The evidence: its weights sum to +1.15, and to +0.46 after the brake.
  3. The counter-search ran, and “nothing found” is recorded as a result.
  4. The opportunity: belief 35% is above the 25% the price assumes, and only the gap is sized.
  5. Risk: its volatility is 25%, so the size is 2.3% instead of 3.8%.

Illustrative examples, not measurements.

Self-improvement

They measure the agent, train it, specialise it and record its work.

  1. Evaluation

    Measures every agent on real tasks: does it find the figure, compute it with a tool, cite it, and stop when it does not know?

  2. Training environments

    Every failure becomes an environment to practise in: real documents and tools, and a question with a checkable answer.

  3. Specialisation

    An open model is trained on one task, then scored against a general model on questions it never saw.

  4. Operation and record

    Runs the agent and records every step, so any decision can be rebuilt from its record.

Illustrative, no values: each drawing shows what a capability does, never a result.

Try it on a task from your work.

In a research partnership: your documents, your experts’ judgement.

Tell us about your taskAbout the partnership