v0.2.0a0 — open source
Simulate how any company
reacts to change
Mimic turns a company's public financial reports into a working model. Ask "what if?" — a price spike, a port closure, a new tariff — and watch how the business would likely respond, before it happens.
how it works
Three steps to a working model
01
Learn
Mimic reads a company's public financial reports, market data, and earnings calls, then builds a working picture of how the business runs — what it sells, who supplies it, and where the money comes from.
02
Simulate
Hit it with a "what if" — a supply shock, a rate hike, a new tariff. AI plays the company's decision-makers, while real financial math keeps every number grounded.
03
Cascade
No company stands alone. One company's move changes the world around it, and its partners and rivals react in turn — so you see the ripple effects across the whole supply chain.
ecosystem
Eight packages. One platform.
Each one works on its own. Together, they show how companies behave under pressure.
mimic
AvailableThe core. Turns a company's public filings into a model you can experiment on.
mimic/world
AvailableConnects many companies at once, so you can watch a shock spread from one to the next.
mimic/memory
AvailableGives every model a memory — so a simulation can be saved, replayed, and trusted to repeat exactly.
mimic/concordia
AvailablePuts rival companies in the same scenario, each making its own decisions, to see how they play off each other.
mimic/bench
AvailableOur report card. We test predictions against 200 real events from the last decade.
mimic/signal
AvailableSpots big events the moment they happen — from news, filings, shipping traffic, and markets.
mimic/sim
AvailableRuns thousands of possible futures to show the full range of what could happen, not just one guess.
mimic/forecast
AvailablePlugs in today's best AI forecasting models to predict what comes next.
what 500 simulations revealed
How long a crisis lasts decides 92.6% of the outcome.
How bad it gets? Just 1.7%.
Most risk models obsess over severity. They may be watching the wrong thing.
interactive demo
Try it yourself
Type commands to explore the API. Try help, twin WMT, or simulate.
from mimic import Twin
twin = Twin.from_ticker("AAPL")
result = twin.simulate(
event="Taiwan conflict",
severity=0.8
)
print(result.financial_impact)benchmark
Validated against reality
Every simulation is tested against 200 historical events. We compare predictions to what actually happened.
200
Historical Events
50
Companies Tested
2,847
Predictions Checked
71.3%
Accuracy
key finding — 500 simulations
How long a crisis lasts explains 92.6% of the outcome.
How bad it gets explains just 1.7%.
Most risk models focus on severity. Our data says that's the wrong thing to watch.
$ mimic benchmark --events=crisis_2015_2024
Running benchmark on 50 S&P 500 companies...
Event: COVID-19 pandemic onset (2020-03)
WMT predicted: "Accelerate e-commerce, stockpile essentials"
WMT actual: Matched (e-commerce +74% YoY)
Score: 0.84
Overall Results:
Accuracy: 71.3% across 2,847 test pairs
the math behind the numbers
dcf_impact()EV change from cash flow shockcogs_sensitivity()Input cost → margin impactaltman_z()Bankruptcy risk scoresupplier_hhi()Supply concentration (HHI)capm_response()Stock reaction to marketcascade_propagate()Supply chain propagation