From 178 Orgs to 19 Proposals

This practice exists because I spent time inside the Project Liberty ecosystem.

I was at Project Liberty Labs building Frequency — the Layer 1 blockchain that implements DSNP, Frank McCourt's decentralized social networking protocol. That's where I worked with Lara and the team, and that's where I saw the broader network up close: the Project Liberty Alliance, 175+ organizations all working on some version of building a better internet, protecting digital rights, or strengthening civil society.

When I started thinking about where AI/ML infrastructure could create the most value for mission-driven organizations, I didn't have to look far. The network was already there.

The Research Sprint

178 organizations in the PLA ecosystem. All doing some version of "fighting the hardest problems on earth with inadequate tools." The question was which ones would benefit most from fractional AI/ML engineering — and the only way to answer that was data, not intuition.

Every org in the network got scraped, enriched, and scored. Not a casual scan — a systematic analysis across six dimensions: AI/ML relevance, tech gap, impact potential, size fit, build viability, and growth trajectory. Each dimension rated 1-5, weighted by what actually predicts whether fractional engineering creates value.

The enrichment layer was the most revealing part. You can't score an organization's tech gap from their website alone. You need to know: what tools do they actually use? Do they have any engineering staff? What does their daily workflow look like? Are they already getting help from Google's or Microsoft's in-kind programs?

Some of this came from public data — job postings, annual reports, tech stack signals. Some came from deeper research. The goal was to understand each organization well enough to know whether a proposal would be worth writing.

What the Data Showed

The 178 orgs sorted themselves into clear tiers:

High fit (tech gap 4-5, 10-80 people): Organizations doing data-intensive work entirely with manual processes. Policy researchers reading legislation by hand across 50 states. Investigative journalists processing Telegram channels one message at a time. Child safety advocates screenshotting dark patterns in apps individually. These orgs have deep domain expertise and critical missions — they just don't have anyone building infrastructure for them.

Medium fit (tech gap 3, or size edge cases): Orgs that have some technical capacity but could use targeted help. These might become projects later, but they're not the first priority.

Not a fit: Organizations that already have engineering teams, are too large to benefit from fractional work, or whose mission doesn't involve data at a scale where ML creates leverage. No shame in it — they just don't need what we build.

From Scores to Proposals

The top-scoring organizations became project proposals. But the scoring only got us to "this org is a good fit." The proposal itself required a second round of deeper research: what does this team actually do every day? Where is the specific bottleneck? What would we actually build, and in what order?

Each proposal on this site follows the same structure: the opportunity, the problem today, what we'd build, and the build phases. Every claim is grounded in what the org actually uses and does — not generic AI pitches. When we say IST tracks 48 policy recommendations across 50 states by hand, it's because their annual report says so. When we say CyberPeace Institute classifies threat reports manually, it's because we researched their workflow.

19 proposals. 6 domains. Press freedom, child safety, cybersecurity, climate justice, open knowledge, and humanitarian tech. Each one a concrete, buildable project scoped for a real organization.

What Happens Next

These proposals are just the surface.

Everything on this site is based on external research — public data, annual reports, tech stack signals. It's informed, but it's still an outsider's view. The real work starts when we get inside an organization and validate these use cases against what's actually happening day-to-day.

That's where it gets interesting. Every org we've researched has the problems we've scoped in their proposal. But they also have ten other things we can't see from the outside — the broken CRM integration, the grant reporting process that eats two weeks every quarter, the volunteer coordination that runs on group texts, the donor data that lives in three different spreadsheets nobody trusts. The proposals are the first layer. Engagement peels back every other layer underneath.

The model is: start with a validated use case, then expand as we understand the full picture. The proposal gets us in the door with a concrete, buildable project. Once we're embedded and building, we see everything else — and the scope grows naturally from there.

Lara and the PLA team provide the warm introductions. The research provides the credibility. The proposals provide the starting point. From there, it's about peeling back the onion and building what actually matters.