How the examples were chosen & analyzed

A documented, reproducible funnel — not cherry-picking. See the example reports.

Sampling

We sample public GitHub repositories across five product domains and three builder-experience tiers (by the repo owner's public follower count), drawing roughly evenly across the grid with one repo per owner and a fixed random seed (seed=42) for reproducibility. Each candidate is shallow-cloned and passed through a free product-like gate — we only spend on repos that read as real, shippable apps (a deploy config, a live site, or an actual screen/route surface), dropping libraries, CLIs, and config repos.

Tiers by owner followers: Newcomer (0–25), Emerging (25–250), Established (250–∞).

What was sampled

DomainSearch queryEvaluated
Games topic:game fork:false archived:false pushed:>2024-06-20 size:50..200000 19
Consumer productivity topic:productivity fork:false archived:false pushed:>2024-06-21 size:50..200000 12
DevOps topic:devops fork:false archived:false pushed:>2024-06-21 size:50..200000 15
Finance topic:fintech fork:false archived:false pushed:>2024-06-21 size:50..200000 21
Education topic:education fork:false archived:false pushed:>2024-06-21 size:50..200000 10

Sampled 2026-06-10. GitHub search results drift over time; the seed + date make a run reproducible.

The analysis

For each project Brix derives a plain-language product brief from the code, then researches the live market: real competitors and buyer personas (how many, what they'd pay). It then runs a persona × build-level simulation — for four build levels (today's MVP through category leader), it models which personas would choose this product over the competitors they already use, and turns that into an addressable audience and a risk-discounted revenue estimate. The headline Brix Value is that revenue minus the cost of the build time. Reports carry citations — treat them as leads to verify, not settled fact.