You're pledging to donate if the project hits its minimum goal and gets approved. If not, your funds will be returned.
$50,000 left to reach the funding goal.
I live in Knightdale, North Carolina. The schools here are starting to use AI for grading and behavior tracking. Small businesses are using it to screen applicants. Parents are told "the AI decided" and have no idea what questions to ask. That is the gap I am trying to close.
I am building a free, open AI Stewardship curriculum for the people who will actually live with these systems: parents, school staff, and local employers. It is five modules, about 24 hours total. The core is a simple four-step protocol—verify the output, check the stakes, document the decision, and know when to escalate a concern to a school, a clinic, or an employer. The first pilot will run here in Knightdale with 10–15 participants at the public library.
This is AI safety work because safety does not stop at the model. If the people affected by an AI system cannot recognize when it is failing, technical guarantees do not matter. I am building a small, local network of people who can act as an early-warning system for real-world failures. That is field-building, and it is a public good—no company is incentivized to build it.
What the money buys: I am asking for **$20,000** to run the pilot and publish the curriculum under a Creative Commons license. The budget is printing and materials ($2,000), facilitator stipends for two co-teachers ($6,000), participant support for childcare and transportation so working parents can attend ($3,000), evaluation design and data collection ($4,000), and a $5,000 operating reserve. I do not know yet whether the 24-hour format is right. The pilot will tell me, and the evaluation report will be public either way.
Two goals. First, build and pilot a free AI Stewardship curriculum with 10–15 people in Knightdale. Second, publish the whole thing—modules, facilitator guide, and evaluation report—under a Creative Commons license so any library, church, or workforce center can run it without paying us.
I get there in four steps. I finish writing the five modules and the facilitator guide, which I have already outlined. I recruit the first cohort through the Knightdale public library and two local partner organizations. I run the pilot over six to eight weeks with two paid co-facilitators, so it is not one person improvising. Then I publish the results, including what did not work.
The thing I am actually building is not a course. It is a small network of people who know how to check an AI system and know who to tell when it is wrong.
Three measures, and I will report all three whether they look good or not. Knowledge gain: a scenario-based assessment before Module 1 and after Module 5. Not "define AI safety"—real prompts like "here is an AI output, is it credible, what would you check." I report the percentage-point gain. Demonstrated competence: at the end, each participant gets a realistic scenario—an AI system ranking job applicants, or a school using AI for behavior scoring—and has to apply the four-step protocol: verify, contextualize, document, escalate. I score it on a simple rubric. Behavior at 90 days: a follow-up survey asking whether they used the protocol, on what, and whether they raised a concern with an institution. This is the measure I am least confident about. Short-term learning gains from programs like this are well documented; long-term behavior change is not. The 90-day follow-up exists to find out, and I will publish whatever it says.
$20,000, itemized:
· Printing and materials — $2,000. Workbooks for participants without reliable internet at home.
· Facilitator stipends, two co-teachers — $6,000. Paid, not volunteer, so the pilot is run properly.
· Participant support, childcare and transportation — $3,000. The biggest reason working parents in Knightdale drop out of evening programs.
· Evaluation design and data collection — $4,000. Pre/post assessment, rubric scoring, 90-day follow-up, written report.
· Operating reserve — $5,000. Covers what I got wrong in the budget, which will be something.
Manifund's 5% fee is paid by donors on top of this, so none of the $20,000 goes to platform costs.
I started Keep A Vision Leadership Academy, which works with fatherless boys at risk of gang recruitment and incarceration. That work is where I learned the part of this project that is genuinely hard: getting people who have been failed by institutions to sit in a room and trust the person teaching them. Most AI literacy programs fail on that, not on the content. The AI safety content is the easier half, and I am not building it alone — a technical reviewer is being recruited now and will be named here once they have agreed. I am in federal custody until April 2027, when I go home to Knightdale. That constraint is why this is a $20,000 pilot and not a national program, and it is why the budget is built around what I can finish now: the five modules, the facilitator guide, the evaluation instruments, and signed agreements with the library and two partner organizations. The in-person cohort runs the month I am home. I would rather hand you a realistic calendar than an impressive one. My background is youth leadership development, not AI research. I am not going to pretend otherwise. What I bring is a working curriculum outline, a partner path in Knightdale, and a plan for exactly what $20,000 produces. The technical credibility comes from the reviewer and the published evaluation, not from me claiming expertise.
Four ways this goes wrong, honestly ranked.
Most likely: I cannot recruit 10–15 people for 24 hours of evening sessions. Attendance in community programs is brutal, especially for working parents. If I get six people, the evaluation data is too thin to mean anything. Mitigation: I am running it at the library on Tuesday evenings with childcare and transport covered, and I am recruiting through organizations that already have relationships with these families rather than advertising to strangers.
Second: the curriculum is too technical or too soft. If it is too technical, people disengage by Module 2. If it is too soft, it becomes generic "be careful with AI" content that does not change behavior. Mitigation: the pilot itself is the test, and I will rewrite the modules based on where people got lost.
Third: the 90-day follow-up shows no behavior change. People learn the protocol and never use it. That is a genuine possible outcome. It would mean the curriculum teaches knowledge without creating stewards, and the honest response is to redesign around practice and reinforcement rather than more content.
Fourth: I fail as a first-time founder. I have never run a program like this. If the pilot is badly run, the problem is me, not the concept.
The failure mode I am least worried about is the published curriculum being ignored. Open educational resources with real evaluation data behind them tend to get picked up. The risk is that the data is too weak to be worth picking up.
Nothing. $0.00.
This would be the first funding for Keep A.I. Vision. I have not received grants, donations, or outside investment in the last 12 months, and I have not run a fundraising campaign. I am applying to Manifund first because it is a small, fast, first check that lets me run a real pilot instead of describing one.
If it helps you calibrate: I am not coming to you with momentum. I am coming with a written curriculum, a local partner path, and a specific plan for what $20,000 produces.