Newton & Carter Labs

AI that has to earn trust, tested in the open.

Newton & Carter Labs is the research, evaluation, and prototyping division of Newton & Carter LLC. The lab designs, builds, and stress-tests AI systems that humans have to trust with real decisions — evaluated against published protocols, on synthetic data, in the open.

The lab works in four functions: human-machine teaming research — how operators calibrate trust in AI recommendations, and which interface patterns fix over-trust and under-trust; AI evaluation and testing — accuracy, robustness, failure-mode, and human-factors evaluation run against written protocols; rapid prototyping on synthetic data, so decision-support concepts can be demonstrated where real data is restricted; and software test and IV&V support, offering the same evaluation discipline as a service.

The lab is founder-led: the Founder directs the programs, designs the protocols, and owns every result. Delivery runs through the firm’s on-demand network of 300+ vetted contract specialists, engaged per project. Protocols, methods notes, and sample artifacts are published — the lab’s credibility is checkable work, not claims.

Programs

Three programs of work

Numbered and named like a research portfolio — because that is what they are. Each program produces public artifacts as it advances.

Glass-Box Decision SupportAn interactive prototype and scenario engine for AI-assisted operations management. Every recommendation arrives as a small set of candidate courses of action — each with its rationale, stated confidence, key assumptions, and predicted effects — with a provenance strip on every machine-generated element and automation the operator can see and bound. The design goal is calibrated trust, not maximum trust.Interactive prototype in active development

AI Evaluation & Testing HarnessA documented, repeatable evaluation method for decision-support AI: calibration measurement, failure injection — including deliberately wrong recommendations — explanation-quality scoring, provenance and staleness checks, and override and audit-trail verification. It is the lab’s own quality gate, and the basis of its test and IV&V support services.Methodology v1 published

Synthetic Data & Scenario GenerationSynthetic datasets and operational scenarios — events, sensor feeds, decision injects — for prototyping and evaluation where real data is restricted. Generation is model-based and reproducible, ground truth is recorded separately from what an operator sees, and every artifact is labeled synthetic. Synthetic data is never presented as real data.Methods note published

LAB-002 · Evaluation methodology

How the lab evaluates

The lab’s evaluation protocol (Methodology v1) treats the system-plus-interface as the unit of analysis, in vocabulary aligned to the NIST AI Risk Management Framework and TEVV practice — methodological alignment only; no certification is claimed or implied. Every evaluation checks:

Calibration measurement — stated confidence vs. observed correctness, and operator trust vs. actual system performance
Injected-error detection — deliberately wrong recommendations, catchable from the system’s own displayed evidence
Explanation-quality scoring — rationale, confidence honesty, assumptions, provenance, actionability
Provenance & staleness checks — source, model and version, timestamp on every machine element
Override & audit-trail verification — one-action rejection, bounded automation, a complete and reconcilable log

Findings are reported in full, including the unflattering ones — the protocol’s first end-to-end exercise ran against the lab’s own LAB-001 prototype, and its sample report, majors and all, is the lab’s published format reference.

Division facts

About the lab

Division of: Newton & Carter LLC — a Texas limited liability company (single-member), formed October 2026, Leander, Texas. Contracting runs under the legal name Newton & Carter LLC.
Directed by: Chris Carter, Founder, Newton & Carter LLC
Focus areas: Human-machine teaming · AI evaluation & testing · Synthetic data & scenario generation · Test & IV&V support
SAM.gov registration: In progress — UEI: [UEI pending] · CAGE: [CAGE pending]
Contact: chris@newtonandcarter.com · newtonandcarter.com/labs

Have a system that needs evaluating?

Test and evaluation inquiries, teaming, and research collaboration — one conversation starts it.

Start a conversation