01
Coordination is the real bottleneck
Tax and benefit policy is still interpreted through closed, non-interoperable models. Institutions do not just disagree on values. They often are not running the same system at all.
Society, in silico.
APIs for household calculation, rule lookup, microsimulation, and imputation.
Cosilico exposes household calculation, rule lookup, microsimulation, and imputation through a managed runtime for economic products, policy tools, and AI systems. The stack is open source; Cosilico offers the managed runtime.
Open-source stackDeterministic rule executionCalibrated household simulation
System sketch
Rule libraries
Executable policy logic assembled from open-source rule encodings and shared runtime components.
Rule substrateValidation
Parse checks, references, example tests, and external comparisons.
VerificationPopulation model
microplex and statistical matching build calibrated synthetic households.
Household modelCosilico APIs
Household calculation, rule lookup, microsimulation, and imputation through a managed runtime.
Managed runtimeWhy this matters
Most institutions are not arguing from the same model.
The thesis is not that policy needs prettier charts. It is that rule interpretation, household data, and simulation should run on a common substrate instead of being rebuilt inside every institution.
02
The missing layer is shared runtime
Calculators exist. Forecasts exist. Scorekeepers exist. What does not yet exist in the open is a queryable, household-level runtime that connects encoded law, calibrated populations, and simulation at operational scale.
03
Open infrastructure should carry the trust burden
The legal and modeling substrate should be inspectable. Cosilico can leverage open encodings and source infrastructure without making them the center of the product story.
Proof
Trust comes from artifacts, not posture.
The public trust story should start with the evidence already present in the repo: published validation work and concrete structural analysis.
4/4
published test cases matched
Current public validation artifact: 26 USC §1 2018 TCJA rate schedules matched expected outputs.
100%
accuracy on that artifact
The published report records full agreement across the tested filing-status schedules.
$0.01
comparison precision
The validation methodology compares outputs to cent-level tolerance rather than using hand-wavy qualitative checks.
109k
cross-references mapped
The statute-graph analysis reports roughly 109,000 cross-references across 6,680 Internal Revenue Code subsections.
Current public proof points come from the published 26 USC §1 validation report and the statute-graph analysis artifacts in the repo.
Stack and runtime
Everything is open source. The offering is the managed runtime.
Stack
Rule libraries
Axiom Foundation encodings and related open rule libraries that Cosilico can leverage as supporting substrate.
axiom-foundation.orgmicroplex
Open-source synthetic microdata synthesis and survey reweighting for calibrated household populations.
GitHubpy-statmatch
Open-source statistical matching and data-fusion tooling for combining partial datasets into richer analytical records.
GitHubValidation artifacts
Published reports and model checks are part of the trust surface, not internal-only ops debris.
Managed runtime
Managed APIs
Managed endpoints for household calculation, rule lookup, microsimulation, and imputation with authentication, billing, and routing.
Fresh calibrated data
Operational datasets, versioned inputs, and durable query surfaces so downstream teams do not have to reassemble the stack themselves.
Elastic compute
Population-scale simulation and batch workloads that are expensive, operationally messy, or too brittle for each user to self-host.
Runtime operations
Access control, credits, observability, and developer ergonomics belong in the managed runtime, not in hidden model logic.
Architecture
Law, populations, and compute in one pipeline.
The product is not a single model. It is the path from encoded source law to repeatable answers, with validation and calibrated household data in the middle.
Source and encode law
Open rule libraries capture statutes, regulations, and guidance as machine-readable logic with provenance back to source text.
Validate rule behavior
Encodings are checked with parse validation, reference integrity, test-driven examples, and comparisons against external systems where available.
Build synthetic populations
microplex and py-statmatch create calibrated, privacy-preserving household populations and fused records that can support counterfactual simulation.
Serve queries and compute
Cosilico exposes the runtime through hosted APIs for household calculation, rule lookup, microsimulation, and imputation.
API surfaces
A concrete offering with room to expand.
The product story becomes credible when it resolves into actual interfaces instead of abstract platform language.
/calculate
/calculate
Deterministic household tax and benefit calculations with an audit trail back to encoded rule logic.
/rules
/rules
Structured lookup into encoded rules, thresholds, applicability, citations, and dependencies for retrieval and agent workflows.
/microsim
/microsim
Population-scale reform analysis, distributional breakdowns, and scoring runs over calibrated households.
/impute
/impute
Complete partial records with statistically grounded attribute prediction and calibration-aware enrichment.
Who needs it first
Start where deterministic economic reasoning already matters.
Policy teams
Teams that need distributional analysis, scoring, and explicit assumptions instead of one-off spreadsheet logic.
Financial institutions
Teams that care how policy and household cash flow propagate into defaults, spending, portfolio risk, or segmentation.
AI systems
Agents and applications that need a deterministic economic reasoning backend instead of free-form textual guesses about rules.
“
The ambition is large. The trust has to be earned artifact by artifact.