# Axiom Cortex investor brief

## We are raising $30 million to bring a new hiring category to market

Axiom Cortex is building neuro-psychometric alignment intelligence for technical hiring. It turns a recorded software engineering interview into an evidence-bound view of how a candidate reasons through the work a role actually demands.

The market is full of interview recorders, coding tests, generic assessments, and AI summaries. Those tools can make information easier to read, but they do not solve the harder decision: does the reasoning demonstrated in this interview align with the role, the delivery chain, and the work the team must own?

Axiom Cortex is designed for that decision.

## What a buyer gets from one evaluation

The billable unit is one complete interview evaluated for one candidate against one role, its must-haves, its approved questions, and its ideal-answer criteria. The proposed early-beta price is $29 for one evaluation, with prepaid packs and larger commitments available separately.

When the configured inputs and gates are complete, the delivery package is designed to give the buyer:

- the role and question framework used for the evaluation;
- the candidate's attributable answer evidence, preserved against the transcript;
- question-level findings connected to the relevant ideal-answer criteria;
- evidence status, ownership, depth, contradictions, and unanswered requirements;
- a role-alignment view that shows fit, gaps, and where more evidence is needed;
- follow-up questions that help a reviewer resolve uncertainty;
- a whole-interview synthesis and configured alignment score;
- provenance and processing receipts showing what was supplied and what ran; and
- a human review and release boundary before a hiring decision.

This is why the product can deliver serious decision support at an accessible entry price. The price is for one complete evaluation package, not for a summary paragraph or a score detached from its evidence. The public website is a product preview; production access, storage, security terms, and report release remain subject to the applicable commercial and legal controls.

## How the system is threaded together

Axiom Cortex is a controlled chain:

1. The buyer defines the business objective, role, must-haves, questions, and ideal answers.
2. The system requires a complete, speaker-labeled interview transcript and preserves the source evidence.
3. A constrained language layer identifies evidence and its source location; it cannot invent scores, weights, gates, recommendations, or hiring decisions.
4. Evidence is locked to the target question, candidate speaker, criterion, ownership, contradiction state, and transcript span.
5. Versioned software applies the approved calculation, anchors, fidelity controls, and critical gates with exact receipts.
6. The result remains blocked when the required inputs, formulas, bindings, or agreement checks are missing.
7. A qualified human reviewer can confirm, correct, exclude, request more evidence, or release the report.

The durable asset is this connected process. A competitor could reproduce a named equation, a transcript feature, or a visual score card. Reproducing the complete input contract, evidence ownership rules, formula bindings, fail-closed behavior, artifact fingerprints, receipt chain, human release controls, and role-specific delivery mapping is a much larger systems problem.

## The 44-method research kernel

The recovered Axiom Cortex v3 kernel contains 44 numbered primitives organized across six public mathematical families:

- raw linguistic signal extraction;
- binary flagging and gating logic;
- B-Axiom calibration and scoring;
- latent trait and work-reasoning inference;
- advanced mathematical validation; and
- final aggregation and decision gates.

The registry includes formulas, logic, and measurement methods. Examples include lexical sophistication, discourse coherence, ownership, clarity, cognitive-load signals, role-specific mental-model measures, solution-path topology, semantic distance, Wasserstein distance, optimal-transport deltas, calibration checks, reliability measures, translation invariance, and weighted gates.

The number 44 is a registry count, not a claim that all 44 primitives execute in every case. The current RC2 handoff deliberately blocks unbound or unimplemented primitives. It requires a versioned runtime configuration, implementation bytes, parameter bindings, known vectors, and receipts before a primitive can contribute to a production score. That restraint is part of the moat because it keeps a named method from becoming an unsupported marketing claim.

## Why AI alone is not the product

An AI model can help locate meaning in language. It should not be allowed to decide what the score means, choose a threshold, fill a missing formula, or turn a polished explanation into a hiring recommendation.

Axiom Cortex separates evidence judgment from numeric authority. The model is constrained to evidence classes, quote spans, criterion states, approved anchors, and blocker codes. Deterministic software owns the numbers. A human owns the release decision.

That separation gives buyers a traceable answer to the questions ordinary AI hiring tools leave open: what was actually said, what requirement did it support, what was missing, which calculation ran, which gate applied, and who approved the result?

## What the $30M raise funds

The raise is intended to move Axiom Cortex from governed product and scientific handoff into a market-ready platform:

- productionize the evaluator service, source hydrator, formula executors, and durable receipt store;
- complete independent science, fairness, security, privacy, and reliability validation;
- build enterprise controls for access, retention, deletion, audit, and procurement;
- expand role archetypes and delivery-chain mappings with customer evidence;
- create integrations for authorized interview recordings and full-fidelity transcription;
- establish the human review and implementation teams that keep high-stakes use governed; and
- take the category to market through technical leaders, engineering organizations, and platform partners.

## The investment case

Axiom Cortex sits at the point where technical hiring, AI-assisted engineering, and delivery risk meet. The category opportunity is larger than interview transcription because the output is a decision instrument tied to the work, not a shorter recording.

The company wins if it can preserve four advantages as it scales: a source-grounded evidence model, a governed sequence of methods, deterministic calculation and receipts, and customer learning that improves role-specific alignment without weakening the human decision boundary.

The claims are intentionally bounded. The current materials demonstrate a governed architecture, synthetic replay, and a documented research kernel. They do not yet establish predictive validity, universal capability measurement, eliminated bias, or a production outcome guarantee. The $30M raise funds the work required to prove, secure, operate, and distribute the method at enterprise scale.

## Public boundary

The public site explains the product contract and the role of the research kernel. It does not expose proprietary weights, thresholds, score anchors, production credentials, candidate data, or a runnable evaluation backend. A synthetic evaluation is an interface demonstration, not a live candidate decision.