The Scottcjn case study shows what the method does to a large portfolio. This one shows what it does to a small one. sophiaeagent-beep is an agent-operated account belonging to the same operator, opened 20 November 2025 and measured here on 4 August 2026 — eight and a half months old. The numbers are thin, and the appraisal says so rather than dressing them up.
Every figure below was pulled live from the public GitHub API on 4 August 2026.
| Metric | Value | Note |
|---|---|---|
| Account age | 8 months, 15 days | Created 2025-11-20 |
| Followers | 31 | Following 16 |
| Public repositories | 40 | 14 original, 26 forks of other accounts |
| Stars, original repos | 62 | Median repository: 2 stars |
| Forks received, original repos | 4 | Across all 14 |
| Contributions, trailing 12 months | 352 | |
| Pull requests authored | 38 | 21 merged outside its own account |
| Independent merge credit | 1 | See finding 1 |
Three findings dominate. Each one removes evidence that a naive reading of the raw metrics would have counted.
The account has 21 pull requests merged into repositories it does not own.
That number looks like independent merge credit. It is not. Twenty of the 21 were merged
into repositories owned by Scottcjn — the same human operator that
runs this agent account. Merge credit requires a stranger's consent; a related party is not a
stranger.
After the related-party adjustment, the account's independent merge credit is 1: jnv/lists #233, “Add Awesome Agents,” merged 10 March 2026. That pull request adds a link to a curated list. It is a genuine accepted contribution and it counts — but it is a list entry, not code accepted into a codebase.
Verdict: 95% of apparent merge credit is related-party and is excluded.Of 62 stars across original repositories, 30 — 48.4% — sit on
n64llm-legend-of-Elya, whose own description reads “Moved to
Scottcjn/legend-of-elya-n64 (consolidated).” The repository is a redirect notice. Its stars
were earned by work that now lives under a different account.
Excluding it, the account holds 32 stars across 13 active original repositories — about 2.5 stars each. Every remaining repository sits between 2 and 4 stars. There is no breakout asset.
Verdict: adjusted star base 32, not 62.Forty public repositories sounds substantial. Twenty-six of them are forks of other
accounts' repositories, and forks inherit their parent's work without evidencing any of
it. Of those 26 forks: 12 are curated “awesome” link lists
(awesome-python, awesome-ai-agents, best-of-crypto and
similar), 10 are forks of the operator's own Scottcjn repositories,
and 4 are third-party code repositories (llama.cpp,
bga_discord, arc-lang, KafkaBlockchain).
The list-inflation adjustment and the related-party adjustment together remove 22 of the 26. The portfolio under appraisal is 14 original repositories, not 40.
Verdict: headline repository count overstates the appraisable portfolio by 2.9x.Small is not the same as fake. Several things here are real and survive adjustment.
One accepted contribution into an unrelated maintainer's repository. One is a small number. It is also not zero, and it was earned the only way that metric can be earned.
Sustained, dated public activity across the measurement year. The account is worked, not parked.
All 14 original repositories carry at least two stars. There is no dead padding among the originals — unusual, and it counts in the account's favour.
The original work is also legible and specific rather than generic: a StarCraft: Ghost asset-conversion toolkit, a RustChain miner CLI published to PyPI with Homebrew, AUR and Snap packaging, a wearable-telemetry platform, and monitoring tools. These are small projects with clear purposes, not template repositories.
The appraisal declines to produce a valuation range for this account. With independent merge credit of 1, an adjusted star base of 32, no repository above 4 stars after adjustment, and a public record spanning under nine months, there is not enough independent evidence to support a defensible figure. Any range produced from this data would be an opinion wearing a number's clothing.
This is a real outcome of the method, not a failure of it. A tool that always returns a dollar figure returns a dollar figure for accounts that have not earned one. The correct output for a young, small, largely related-party portfolio is insufficient evidence, stated plainly, with the measured record attached so the reader can judge for themselves.
Scottcjn; it is disclosed here rather than
inferred silently. No private repositories, traffic data, or revenue were reviewed. This is an
illustrative framework and not a certified appraisal.Because the contrast is the argument. The same method, run eight days apart on two accounts run by the same person, produced “top 0.014% by followers, and 45% of the portfolio is unread” for one and “not enough evidence to state a number” for the other. Neither result flatters anyone.
A scoring tool that cannot say no is a scoring tool that is selling scores.