Problem
Finding a co-founder usually happens through warm intros, college circles, or crowded startup communities. That works if you already know the right people. It breaks down for early builders who have intent but no discovery layer for skills, goals, and stage fit.
The gap: founder discovery needed to feel more like product matching than social networking, with less noise, faster qualification, and clearer signals about who should actually talk.
Users
Student builders and first-time founders looking for a collaborator with complementary skills. The working assumptions (not a formal study) were:
- Capability, seriousness, domain interest and build stage matter more than a polished bio.
- People would sign up only if discovery felt fast and the first useful match arrived quickly.
Decision
Trade-offs
We made three product calls for the first version:
- Speed over completeness. A light profile now, richer signals later.
- Legible matching over a black-box score. Users needed to understand why a profile surfaced, so discovery leans on a small number of visible inputs (skills, domain expertise and location filters) instead of an opaque ranking.
- Treat it as a learning build. The first version was an MVP to learn from, not a finished network.
What shipped
- Skill-first profiles. Profiles centre on skills, domain expertise and what the person wants to build, pushing users toward intent-rich profiles instead of vague networking.
- Explore flow. A quick yes/no pass through profiles, with filters, instead of long directory browsing.
- Mutual matches. A like on both sides creates a match.
- Four core surfaces: onboarding, explore, profile and matches.
Validation
There is no usage evidence to report. Hackmate was shared with peers, but usage was never instrumented, so there is no user count, activation rate or match rate, and no user numbers are claimed here.
Limits and next
- No instrumentation. The first thing to add would be post-match quality signals: conversation outcomes, follow-up prompts, and whether a match led to collaboration.
- Not re-run. The app was not re-run for this write-up; only the public repo's file tree was checked. This page describes the product design, not a current deployment, and usage was never instrumented, so no user numbers are claimed.
- What I'd change: marketplace quality compounds faster than raw signups, so I would measure match quality before growth.
Credits
A two-person build with a partner; my part was product. Source: hackmate-rework (public repo; returned HTTP 200 on 30 Sep 2026).
