Tausch Capital Labs

Five models.
One human in the loop.

Tausch Capital Labs builds the intelligent investment systems behind the Marylin index. Each model has one job: score the universe, challenge a manual trade, judge a pitch, improve its own playbook, evaluate and compare portfolio performance.

Be useful for investing

Human in the loop

Chris

The only human in the lab

Currently, not all human capabilities can be reliably replicated by smart AI+harness systems. We keep the human in the loop on purpose — for now.

The models

01

Marylin

Our frontier investment intelligence system

Scores a global universe of companies each month across 31 dimensions, aggregates them into five composite factors, and turns the ranking into portfolio weights. The result runs live and public as a wikifolio index.

Live · wikifolio
02

Manni

A critical reviewer of human portfolio adjustments

Takes a proposed manual trade in plain language, restates it as a canonical record, and computes post-trade weights, funding balance and concentration change. It then researches both legs and reports what the proposal gets wrong. Decision support, with execution left to the human.

In production
03

Wanda

An agentic, human-in-the-loop judge for stock pitches

Reads a written pitch, argues both sides, and returns a verdict.

In testing
04

Will

A self-improving investment system willing to outperform

Reviews its own results each day and proposes edits to its own playbook.

Running daily
05

Bengi

A comprehensive benchmarking system

Evaluates and compares the performance of investment portfolios.

In development

The research

The market is the one eval you can’t reward-hack; L($) = the most expensive loss function.

  1. Scalable experiments

    Win by running more high-quality experiments per unit time.

  2. Continuous learning

    The moat is the loop that turns every outcome into a better next decision.

  3. No human in the loop

    Autonomy starts when the system can recognize, contain, and stop its own failures.

  4. Autonomous organizations

    When agents research, execute, and review, the company becomes mechanism design.

All to answer one question:
Can we use AI to beat the market?
Back to reality