Theory of Change

1Purpose and Scope

This document states why LinkScape exists, the long-term change we are working toward, and the logic that connects our day-to-day activities to that change. It is the reference point for every impact document that follows: I-02 Impact Measurement Framework operationalizes the outcomes named here, I-04 Program Evaluation tests whether individual programs move them, and I-07 Annual Impact Report narrates progress against them. Where this document and a program plan disagree about intended outcomes, this document governs until formally amended.

LinkScape is a youth-led open-source and AI nonprofit, founded December 19, 2022 by Thomas Wu. We are a fiscally sponsored project of The Hack Foundation dba Hack Club, a 501(c)(3); we are not a separate legal entity, and all funds flow through Hack Club Bank (HCB). Our team is roughly ten students from CMU, MIT, and high schools, some of whom are minors. That composition is not incidental to our theory of change: we are simultaneously the people doing the work and part of the population we aim to reach.

This is a living strategy document. It is reviewed annually alongside G-13 Strategic Planning and may be amended only by the Founder and Co-founders collectively, consistent with the constitutional authority described in G-02 and G-03.

2Mission Foundation and the Three Pillars

Our theory of change is anchored in the mission stated in G-01 Mission, Vision & Values, which organizes our work into three pillars. Every activity we run must serve at least one pillar, and our flagship efforts serve all three.

2.1Building Impact

We build and release useful, open AI artifacts — research, models, tools, and datasets — that lower the cost of learning and doing AI for people with limited access. Open-source is our default distribution mechanism: an artifact that is not shared cannot compound. Our NLP research, submitted to ACL Rolling Review 2027 and pending decision, and our AI-music project are examples of pillar-one output.

2.2Developing Talent

We turn curious students into capable practitioners by giving them real problems, real compute, and real mentorship. The clearest instance is that LinkScape members themselves grow from participants into researchers, maintainers, and competition entrants. Our 80x NVIDIA H100 GPU fleet is what makes this credible — very few youth organizations can offer members genuine training-scale compute.

2.3Fostering Community

We create durable spaces — hackathons, workshops, competition teams, and open repositories — where people who would otherwise work alone collaborate, teach each other, and stay connected. Community is both an outcome we value in itself and the mechanism through which the first two pillars scale beyond our own membership.

3Long-Term Impact Goals

These are the multi-year changes we hold ourselves accountable to. They are deliberately ambitious and are not expected to be fully achieved within any single reporting year; I-02 tracks movement toward them.

• Access: Students and communities with limited prior exposure to AI gain practical, hands-on capability they would not otherwise have had — not just awareness, but the ability to build.
• Contribution: A growing body of open, reusable AI artifacts authored or co-authored by young people exists in the commons and is actively used by others.
• Trajectory: People who pass through LinkScape go on to further AI education, research, employment, or their own community-building — the organization functions as an on-ramp, not a terminus.
• Norms: The practice of responsible, open, youth-led AI is visibly modeled, showing that rigorous and ethical AI work does not require a corporate lab or a graduate degree to begin.

4Activities Mapped to Outcomes

The table below maps each core activity to the pillar it primarily serves, the immediate output it produces, and the outcome it is intended to move. Read it as the connective tissue of Section 6's causal chain.

5Key Assumptions

A theory of change is only as sound as its assumptions. We state ours explicitly so I-04 Program Evaluation can test them and so we can detect early when one is failing.

1. Access unlocks capability: Given real compute (our H100 fleet, governed by F-03), mentorship, and a concrete problem, motivated students can reach genuine competence faster than traditional pathways allow.
2. Open beats closed for our mission: Releasing work openly produces more downstream learning and reuse than any advantage we would gain by keeping it private.
3. Peer-led learning scales: Students learn effectively from slightly-more-experienced peers, so our own members teaching others is a multiplier rather than a dilution of quality.
4. Fiscal sponsorship is sufficient: Operating under Hack Club / HCB (per G-06) gives us the legal, financial, and infrastructural backing we need without a separate entity, so we can spend our energy on programs rather than administration.
5. Safety enables participation: Because some members are minors, credible youth-safety and responsible-AI practices (E-01, S-01, PP-06) are a precondition for growth, not an afterthought — a safety failure would halt impact, not merely dent it.
6. Retention compounds: Community ties formed through our activities keep people engaged long enough for capability and contribution to accumulate.

If any assumption is contradicted by evidence gathered under I-02, the affected activities and this document should be revisited at the next G-13 strategic review rather than left to drift.

6The Theory-of-Change Chain

Read as a directed chain, our logic runs Inputs → Activities → Outputs → Outcomes → Impact. Each stage is a necessary condition for the next, and the assumptions in Section 5 are the links that must hold for the chain to carry weight.

6.1Inputs

• Human: ~10 youth members and advisory fellows (Jett Chen, Evan Luo, Tz-yun Hsiao), plus volunteer time governed by PP-12.
• Compute: 80x NVIDIA H100 GPUs, tiered and logged per F-03 and S-01 — our most significant single resource.
• Infrastructure & backing: Hack Club fiscal sponsorship and HCB financial rails (G-06); collaboration tools — GitHub, Slack, Notion, Zoom, WeChat (O-05).
• Funds: donations and program funding, all flowing through HCB (F-04).

6.2Activities

Hackathons, AI/Kaggle competitions, NLP/AI research, AI workshops, and default open-source release of what we build (see the Section 4 mapping).

6.3Outputs

• Trained participants and workshop attendees.
• Public repositories, models, datasets, and tools.
• Peer-reviewed papers and competition submissions.
• Events held and active community connections formed.

6.4Outcomes

• Short-term: participants gain named, practical AI skills and confidence to build.
• Medium-term: members and alumni contribute reusable open artifacts and progress to further AI study, research, work, or their own initiatives.
• Cross-cutting: a self-sustaining community that continues to teach and collaborate.

6.5Impact

Over years, the accumulation of these outcomes advances the long-term goals in Section 3: broadened access to practical AI, a richer open commons authored by young people, positive trajectories for those we reach, and a visible norm of responsible, open, youth-led AI. This impact statement is what I-07 Annual Impact Report ultimately reports against.

7Boundaries and What We Do Not Claim

To keep our claims honest, we name the limits of this theory of change.

• We contribute to, but do not solely cause, long-term outcomes like a participant's career — many factors outside LinkScape shape those, and I-02 attributes accordingly.
• We do not pursue scale at the expense of youth safety or research integrity; where the two conflict, safety and integrity win (per E-01, E-03, and CFO-owned safety escalation).
• We do not measure success by headcount alone; depth of capability and quality of contribution matter more than raw participation numbers.
• We do not treat compute as an end in itself; the H100 fleet is a means to the outcomes above and is governed as such under F-03.

8Review and Amendment

This document is reviewed at least annually as an input to G-13 Strategic Planning, and whenever a major program change or a contradicted assumption warrants it. Amendments are made by the Founder and Co-founders collectively. Approved changes cascade downstream: I-02 indicators, I-04 evaluation questions, and the I-07 narrative are checked for alignment after any amendment.

Approval and Adoption

Adopted upon signature by the officers below. Pending ratification at Sprint 0 (see LS-REG-001).

Document Maintenance Log