EducateWright launches with $25 million in pledged funding for anti-violence education and active-inference training.

EducateWright, a pre-formation nonprofit initiative formed by SolutionWright and Universal Natural Intelligence (UNI), today announced its public launch alongside a $25 million funding commitment from its founding partners. Its work is plain: anti-violence outreach, education, and the long work of teaching people to teach again. Its belief is plainer: software does not end violence, and code does not teach a child. People do. EducateWright exists to fund those people, and to grow the next ones.

A $25 million funding commitment, pledged outward. The founding partners have pledged $25 million to EducateWright: $20 million earmarked for its education mission and scholarships, and $5 million earmarked for fully-subsidized workforce mentorships across SolutionWright and EducateWright. The value is offered outward, to the learners, interns, and teachers who step forward.

The UNI Workshop.EducateWright’s flagship training is the UNI Workshop, priced at $75,000. It teaches UNI as a buildable active-inference architecture, not LLM tooling. Comparable AI and large-language-model programs commonly price near $30,000; EducateWright optimizes for a durable understanding of UNI and active inference rather than the lowest sticker price. Through the EducateWright grant, the Workshop is eligible for partial-to-full-ride scholarships, so cost is not the gate.

Fully-subsidized workforce mentorships.EducateWright and SolutionWright are opening internships built to fully subsidize the host’s cost of providing each seat: the grant eliminates the financial barrier for the learner, so mentorship and hands-on learning are available to people who would otherwise be shut out. Application processes for both the Workshop scholarships and the subsidized internships are open now.

Honest about the science. UNI is our architecture and framing for building agents from standard active inference and the Free Energy Principle, the established science associated with Karl Friston and colleagues. UNI is not presented as a new scientific field or a new free-energy formalism, and not as achieved general intelligence, consciousness, or biological equivalence. The audit-grade collaborative review this work rests on, An Organic Operator and AI Operator Collaborative Review of Active Inference Free Energy Minimization: Reviewable Foundations, Reproducible Tests, and Open Tensions (Polzin et al., deposited 26 April 2026 on Zenodo under an MIT license, DOI 10.5281/zenodo.19785799), ships with a deterministic reproducibility test suite of 87 pytest assertions across 11 numerical demonstrations, verified on Linux, Windows, and macOS. Layer 1 of the audit, the AI-executable checks, is complete; Layer 2, the human expert review gates, remain pending. Until Layer 2 completes, the work should be treated as a preprint and build-audit artifact, not an authoritative scientific result. Read it, and try to break it.

What we have seen, so far. In UNI simulations, agents configured with minimal engineered priors, weak or uniform priors, or learnable hyperpriors have shown learning and viability-preserving behavior under defined test conditions; agents with learned priors have shown changes in behavior and model organization. These are preliminary observations to be reproduced and documented, not yet peer-reviewed results, and some remain unpublished until they can be presented with adequate methods, tests, and limitations. The results will be published; they will not be withheld.

What we do not claim. EducateWright will not borrow credit it did not earn. It does not claim a tool made a neighborhood safer, that any benchmark is won, or that any outcome is finished. Its role is narrow and honest: fund the humans who do the work, teach the next ones, and keep the receipts where anyone can check them.

A call for organic operators. EducateWright is inviting learners, interns, mentors, educators, reviewers, and partners to step forward, to apply, to teach, to challenge the work, and to help build the safe harbor where this science can be pursued, stabilized, and kept honest. We are excited, and the work needs more hands. Three slow breaths. The way is clear.

Formal formation, governance, entity status, and any named partners will be confirmed and communicated after factual and legal review.

The review runs in two layers. Layer 1, the checks a machine can run, is complete and open: 87 pytest assertions across 11 numerical demonstrations, verified on Linux, Windows, and macOS. Layer 2, the human expert review gates, is open now, and it needs qualified reviewers. Until Layer 2 completes, the work stands as a preprint and build-audit artifact, not an authoritative scientific result.

The record and its reproducibility suite are public under an MIT license: DOI 10.5281/zenodo.19785799. To take a Layer 2 gate, or to bring a lens we have not thought of, join the review. Read it, take a gate, and try to break it.