The constraint is no longer “can we build it?”
It’s who should build it together.
Foundry is an AI-native platform that forms startup teams, not profiles. We model ventures by the capabilities they require — not job titles — and match people, ventures, and complete teams across four levels, with explainable, mutual scoring.
Not a “Tinder for co-founders.” Venture → required capabilities → candidate pool → team formation → validation.
AI collapsed the cost of building. Coordination is now the bottleneck.
As one person can ship what once took a team, the scarce resource shifts from 'builders' to the right allocation of human capability — and to finding the few people worth betting a decade on.
Evidence supporting
- Solo & 2-person founding teams reached record share of YC batches.
- AI-native startups reach revenue with smaller teams and less capital.
- Distribution, GTM and domain expertise now outweigh raw engineering capacity.
- Co-founder matching waitlists (YC) signal massive unmet demand.
Evidence challenging
- Founder chemistry is hard to score — most failures are relational, not capability gaps.
- Trust is built in shared adversity, not from profiles; cold matching has weak retention.
- Marketplace liquidity: density in one niche beats global coverage early.
- Self-reported skills inflate; without proof-of-work, scores are noise.
From venture to validated team — in five moves
Foundry never stops at 'you two should talk.' Every match is routed through a capability gap analysis and an optional founder trial that produces behavioral evidence.
Venture intake
Describe the startup in plain language. AI extracts a structured venture model + required capabilities.
Capability ontology
Requirements map to 35 composable capabilities across 6 families — not job titles.
Gap analysis
Current team coverage vs. required. Critical gaps, redundancy, outsourcable & AI-automatable flagged.
Team formation
Coalition formation maximizes marginal coverage + interpersonal cohesion, penalizes redundancy.
Founder trial
A 72h–7d collaboration produces behavioral evidence that updates the compatibility model.
Four ways to match — and Level 3 is the differentiator
Most co-founder platforms stop at 'is this person like me?' Foundry adds venture-fit ranking, person-to-venture scoring, and the defensible core: which coalition of available people best satisfies this venture?
Venture fit
“Which venture should I join?”
Reverse direction: rank all ventures by how well a given person fits. The 'which startup should I join?' question.
Person ↔ Person
“Who complements me?”
A technical founder needs a GTM co-founder. We score complementarity (gap-filling), commitment, timezone, financial & risk alignment, and working style.
Person ↔ Venture
“Does this person fill what this venture needs?”
A healthcare AI venture needs an ML engineer with clinic experience. We score capability coverage against required capabilities, weighted by importance.
Team ↔ Venture
“Which coalition maximizes this venture's odds?”
Greedy coalition formation: maximize marginal coverage + interpersonal cohesion, penalize redundancy. Returns capability map, gaps, redundancy, team score.
Role vs. capability
A startup doesn't need a 'CTO.' It needs full-stack engineering, AI engineering, infra, product development. Foundry models 35 capabilities — each with AI-automatability, so we can tell you what to hire vs. automate.
“We need a CTO, a CMO, a Head of Sales.” — vague, title-driven, ignores that one person can hold several capabilities and one capability can span several people.
“You need AI engineering (95), B2B sales (80), healthcare domain (90), GTM (70).” — granular, composable, weighted, and AI can flag what to outsource or automate.
Full-stack, frontend, finance ops, UI design — lower the human bar for early teams.
AI engineering, B2B sales, domain expertise, fundraising — irreducibly human.
Three things most matching platforms get wrong
Matching is two-sided
A→B = 94 but B→A = 62 is not a strong match. Foundry computes mutual scores and penalizes asymmetry — one-sided interest fails.
Never an opaque score
Every match returns per-dimension scores, gap-filling reasons, and explicit risks. You see why — capability fit 96, commitment 94, timezone 82, risk 95.
Trial before commitment
We don’t say “become co-founders.” We propose a 72h build sprint or 10 customer interviews — then update compatibility from behavioral evidence.
See it compute a real team
Pick a venture, watch the engine form a team, inspect the capability gap map, and read the explainability.