Public launch

The Axiom Foundation goes public on July 28.

Join the virtual briefing for a first look at the rules we've encoded — and everything that launches with them.

Doors open in
––Days
––Hrs
––Min
––Sec

Computable law for all.

Open, machine-readable encodings of the world's rules — starting with tax and benefit policy. Cited, time-aware, and executable, so anyone can run, audit, or reform them.

  • 26 USC § 24Child Tax Credit
  • 26 USC § 32Earned Income Credit
  • 7 USC § 2017SNAP allotment
  • 26 USC § 1411Net Investment Income Tax
  • 42 USC § 1382Supplemental Security Income
  • 26 USC § 36BACA Premium Tax Credit
  • 26 USC § 63Standard Deduction
  • UKSI 2013/376 reg 22Universal Credit elements
  • 26 USC § 3101OASDI payroll tax
  • 26 USC § 401(k)Cash or deferred arrangements
  • 20 CFR § 416.1110SSI earned-income exclusion
  • 26 USC § 24Child Tax Credit
  • 26 USC § 32Earned Income Credit
  • 7 USC § 2017SNAP allotment
  • 26 USC § 1411Net Investment Income Tax
  • 42 USC § 1382Supplemental Security Income
  • 26 USC § 36BACA Premium Tax Credit
  • 26 USC § 63Standard Deduction
  • UKSI 2013/376 reg 22Universal Credit elements
  • 26 USC § 3101OASDI payroll tax
  • 26 USC § 401(k)Cash or deferred arrangements
  • 20 CFR § 416.1110SSI earned-income exclusion
§I · The gap

Law for the digital era

Every benefit calculator, tax program, and policy assistant has to translate the law from human prose into something a machine can run. Most do it from scratch. Most do it differently. Most do it out of public view, making it hard to verify and harder to fix.

There is no shared layer to point at. No canonical source for what the Earned Income Tax Credit actually computes, or how the SNAP standard medical deduction should treat a 2024 medical expense in Tennessee. The text exists. The closed interpretation does too. The connection between them does not.

The Axiom Foundation publishes that layer — statute by statute, citation by citation as rules as code — in the open, free for anyone to use.

Today
  • Each system reimplements the law
  • Numbers without citations
  • No shared way to verify
Encoded
  • One source of truth
  • Every value cites a statute
  • Anyone can run, audit, or reform
Why now
  • AI needs ground truth
  • People are asking models policy questions
  • There has to be an answer key
§II · What we publish

Two layers, both in the open

Source documents anyone can fetch and cite, plus the encodings that make those rules computable, time-aware, and verifiable.

The source corpus

The primary text, gathered and addressable

We gather national statutes, regulations, agency guidance, and state law from official sources — openly licensed, version-controlled, and addressable by citation. The primary text every encoding points back to.

  • · United States Code (national statutes)
  • · Code of Federal Regulations
  • · IRS revenue procedures, rulings, and notices
  • · State codes and tax law
Encodings

Encoded so anyone can compute them

We turn the same rules into machine-readable form — every value cites a section, every clause is dated, formulas are executable. Compiles to native code; runs anywhere.

  • · Cited — every value traces to a statute
  • · Time-aware — effective dates on every clause
  • · Composable — reform a parameter without rewriting
  • · Verified — cross-checked against independent engines
§III · The encoder

Statutes, encoded automatically. Verified before they ship.

An AI-driven pipeline reads a statute, encodes it section by section, and runs the result against oracles like PolicyEngine and TAXSIM before any human signs off.

axiom — zsh
$ axiom encode "26 USC 32"
 
[axiom] Loading 26 USC 32... 81,247 characters
[axiom] 14 subsections extracted
 
[encode] Wave 1: (a) (b) (c) (d) (f) (h) (i) (j) (m) (n)
[encode] Wave 2: (e) (g) (k) (l) — depends on wave 1
[encode] ████████████████████ 14/14 complete
 
[validate] CI: 14/14 passed
[validate] PolicyEngine: 14/14 match
[validate] TAXSIM: 14/14 match
 
[done] 14 RuleSpec files written to rulespec-us/statutes/26/32/*.yaml
Step 01read

Read

Pull the statute. Walk the subsection tree. Plan the dependency graph between siblings.

Step 02encode

Encode

An agent per subsection drafts the encoding, citing the section it came from. The pipeline logs every conflict and retry.

Step 03verify

Verify

Continuous Integration checks, comparison against independent oracles, reviewer agents that explain any discrepancy.

The encoder logs every decision. Disagreements get explained, not erased.

§IV · What it powers

One encoding. Many places to use it.

Encode a rule once, in the open, and everyone can build on it — correctly, affordably, and in plain sight.

for builders01

Calculators that audit themselves

Tax software, benefit estimators, eligibility tools — all running off the same encoding, all able to point at the statute behind any number.

for AI labs02

Ground truth for AI

People keep asking models policy questions. The Axiom Foundation gives them a key — verifiable answers grounded in actual law, useful for both training and inference.

for analysts03

Reform without rewriting

Change a parameter, re-run the calculation. Compare current law against any proposed amendment without touching the surrounding rules.

for the public04

Government in plain sight

Every value cites its source. Every formula is open. Anyone can read the law, run it, and check that the answer follows.

§Coda · The foundation

Doing the public-interest work

Everything we publish — code, data, and the decisions behind them — is open.

Get in touch — hello@axiom-foundation.org