abtop

Minimalist LLM benchmark for absolute top-k accuracy

A small GitHub repository measuring LLM top-k prediction accuracy surfaced only on Bluesky (rank #1 there, rank #47 in our pipeline). No GitHub, HN, or developer community traction exists.

Why now: Bluesky's nascent tech discourse surface is easily gamed with minimal engagement; low bar for trending on small corpus.

• Only signal is Bluesky rank #1 with weight 0.06 — lowest-weighted source in composite
• Zero presence on GitHub trending, Hacker News, X/Twitter, Reddit, Product Hunt, Dev.to
• Single citation candidate exists (Bluesky search URL), no canonical project page or documentation link
• Pool context shows 181/200 entities are single-source — this is baseline noise, not signal

Considerations: Bluesky's tiny pool (7 rows) and lack of engagement depth makes #1 meaningless. Could be self-promotion, a single viral post, or algorithmic artifact. No code velocity, no issues/PRs, no external validation.

EMERGING SIGNAL · Ignore: Do not track; revisit only if GitHub stars, HN front page, or multiple developer sources independently surface it.

Sources: Bluesky search: graykode/abtop