Unpaid volunteer roles. Adults 18+ only. No housing offered.
Join the Lamp-Keepers of Vesta
A small volunteer study house for building safe, honest AI managers. Online first, Los Angeles.
A project of the Order of Minerva, our research and learning circle. You do not need to share any faith or belief to take part.
MISSION
AI systems are starting to manage real work: schedules, inboxes, money, other software agents. Within a few years many people will have an "AI manager" that acts on their behalf. We want those systems to be:
- Safe: they ask before acting on anything risky, they stop when told, and they can't be talked into harming the person they serve.
- Honest: they say what they did and why, and they admit uncertainty.
- Accountable: every action leaves a readable log a human can check.
- Good for the people they serve: they protect people's time, money, privacy and dignity, and they don't manipulate.
The house does three things:
- Study: read and discuss the research on AI agents, oversight, evaluation and ethics.
- Build: make small, open, testable tools (approval gates, audit logs, evaluation checklists, safe defaults for agent routines).
- Share: publish short, plain-language notes and open templates anyone can use.
We are small on purpose. We would rather finish one careful thing a week than promise a revolution.
DAILY RHYTHM
Everything is voluntary and flexible. The rule is a shared rhythm, not an obligation. Online members can take part in about 20-40 minutes a day.
| Part | What it looks like | Time |
|---|---|---|
| Study | One paper, chapter or case study; write 3 lines on what you learned | 20-30 min |
| Quiet | Phones away, focused thinking or a walk; no feeds | 10-15 min |
| Service | One small helpful act for someone in the group or community (review a teammate's work, answer a newcomer, help a neighbor with a tech problem) | 10 min |
| Building | One small commit, test, write-up or checklist item on a house project | 20-60 min |
| Ledger | At day's end, one line in the shared "Ledger of Small Things": what you finished today | 2 min |
Study
What it looks like
One paper, chapter or case study; write 3 lines on what you learned
Time
20-30 min
Quiet
What it looks like
Phones away, focused thinking or a walk; no feeds
Time
10-15 min
Service
What it looks like
One small helpful act for someone in the group or community (review a teammate's work, answer a newcomer, help a neighbor with a tech problem)
Time
10 min
Building
What it looks like
One small commit, test, write-up or checklist item on a house project
Time
20-60 min
Ledger
What it looks like
At day's end, one line in the shared "Ledger of Small Things": what you finished today
Time
2 min
Weekly: one 60-minute online seminar (paper discussion) and one 60-minute build session. Monthly: an open review of the house's work and finances, shared with every member.
ROLES
| Role | Who we're looking for | Typical time |
|---|---|---|
| Researcher-Professor (volunteer) | Faculty, postdocs, PhD students or industry researchers in AI safety, ML, HCI, security or agent systems. Lead one seminar a month, advise one project, review write-ups. | 2-4 hrs/month |
| Ethicist (volunteer) | Philosophers, law or policy scholars, bioethicists, practitioners in tech ethics or consumer protection. Review projects for harm, consent, privacy and fairness. | 2-4 hrs/month |
| Volunteer Builder | Developers, data people, designers, technical writers, students. Build and test small tools and templates. | 2-6 hrs/week |
| Mentor | Experienced engineers, researchers or career changers who pair with newer builders (adults). | 1-2 hrs/week |
| House roles (later, from members) | Seminar host, archivist (keeps the Ledger and notes), safeguarding lead, treasurer (open books) | elected, rotating |
Researcher-Professor (volunteer)
Who we're looking for
Faculty, postdocs, PhD students or industry researchers in AI safety, ML, HCI, security or agent systems. Lead one seminar a month, advise one project, review write-ups.
Typical time
2-4 hrs/month
Ethicist (volunteer)
Who we're looking for
Philosophers, law or policy scholars, bioethicists, practitioners in tech ethics or consumer protection. Review projects for harm, consent, privacy and fairness.
Typical time
2-4 hrs/month
Volunteer Builder
Who we're looking for
Developers, data people, designers, technical writers, students. Build and test small tools and templates.
Typical time
2-6 hrs/week
Mentor
Who we're looking for
Experienced engineers, researchers or career changers who pair with newer builders (adults).
Typical time
1-2 hrs/week
House roles (later, from members)
Who we're looking for
Seminar host, archivist (keeps the Ledger and notes), safeguarding lead, treasurer (open books)
Typical time
elected, rotating
WHAT YOU GET
- A real, finished portfolio piece: your name on open tools, notes and evaluations (with your permission).
- Weekly seminars led by working researchers, and feedback on your work.
- Mentorship and references for people who show up and finish things.
- A calm, focused peer group with a clear daily rhythm.
- Practice with real AI-agent tooling, evaluation and safety design.
- A say in what the house builds, through the monthly open review.
HONEST TERMS (please read before applying)
- Unpaid and volunteer. No one is paid, and no one pays to join. We will never ask members for donations as a condition of taking part.
- Not accredited. The house is not a university, school, degree or certificate program. It does not grant credit or certificates.
- No housing. No housing, stipend or relocation is offered. If that ever changes, it will be in writing, with a lease and real terms, and it will never be a condition of membership.
- Adults only. All members must be 18 or older.
- Free exit, any time. You can leave at any time, for any reason, in one sentence, with no follow-up pressure. Your own work stays yours.
- No binding vows. Any "rule" or commitment here is a personal, voluntary practice. Nothing you say or sign here is a legal obligation, and nothing restricts your job, relationships, money or other communities.
- Safeguarding policy. We follow a written safeguarding and code-of-conduct policy: no harassment, two-adult rule for any one-on-one meeting outside public channels, a named safeguarding contact outside the house's leadership, and a simple way to report concerns.
- Background checks for anyone working with minors. The house itself is adults only. If any member mentors or teaches minors in an outreach program, they must first pass a background check and follow the outreach partner's child-safety rules.
- No pressure, no targeting vulnerable people. We do not recruit people in crisis or under pressure (for example, in grief, financial distress or a mental-health emergency), and we never use urgency, guilt or isolation to recruit. If someone seems to be in crisis, we point them to proper help (in the US, call or text 988) instead of recruiting them.
- Open books and open doors. Finances (even if $0) and decisions are shared with members monthly. Members are free to talk about the house with anyone.
- Privacy. Applications are seen only by the screening team and deleted on request.
- Not the only way. This is one small project among many good ones. We encourage members to stay active in other labs, communities and groups.