Theory of change.
We publish our theory of change because grant reviewers, partners, and learners deserve to see exactly how we believe our programs will produce the outcomes we claim. Every pilot is paired with an evaluation plan from day one, designed with the partner institution and reviewed by an independent evaluator.
Logic model
Inputs
- Open-weight foundation models (Meta Llama family and other open community models)
- Authoritative public data (BLS, O*NET, IPEDS, public job postings)
- Curriculum and program-design expertise from workforce and adult-education partners
- Grant funding and in-kind compute support
Activities
- Build and maintain three open-source program toolkits
- Recruit and onboard institutional pilot partners
- Train and credential teacher fellows and advisor fellows
- Publish research briefs and reference implementations
Outputs
- Working open-source tools deployed inside partner institutions
- Trained cohorts of educators and advisors using AI ethically
- Public research briefs, datasets, and evaluation results
- Reusable curriculum modules under CC-BY
Outcomes
- Higher rate of credential completion among pilot-cohort learners
- Higher rate of relevant first-job placement
- Reduced time from learner goal to credential or job
- Educators report confidence and ethical clarity using AI in their classrooms
Target outcomes (3-year horizon)
Concrete, falsifiable outcomes against which our board, funders, and the public can evaluate us.
Faster credential-to-job pathways
Reduce the time from a learner deciding on a goal to landing a relevant first job or credential by 30% across pilot cohorts.
AI literacy for educators
Equip 500 instructors at Title I and minority-serving institutions with a working, ethical AI practice within 24 months.
Open infrastructure
Every tool we ship is open source under permissive licenses, so any workforce program can adopt it without per-seat fees.
How we measure.
Every pilot includes (a) a baseline collected before our tool is introduced, (b) a comparison group where feasible, (c) an independent evaluator approved by the partner institution, and (d) public release of de-identified results within 12 months of pilot end.
We do not use engagement metrics (sessions, messages, clicks) as primary outcomes. We use credential completion, job placement, wage gain, and validated educator-confidence surveys.
Why we pre-register.
For each pilot we publish the hypotheses, sample size, and primary outcome before data collection begins, in the spirit of the AEA RCT Registry and Open Science Framework. This prevents us from quietly cherry-picking favorable findings and is increasingly expected by serious philanthropic and government funders.
Read our published research.