← The AI Hype Audit — all 181 verdicts
HOLDS UP “Counts pulled from GitHub today” — the rare listicle where every star count checks out
The claim7 GitHub repos with counts “pulled from GitHub today”: microsoft/markitdown 169k (any document → clean markdown), ponytail 90k (makes AI write LESS code), awesome-mcp-servers 91k, crewAI 56k, goose 51k, skyvern 22k, mcp-use 10k — “number 1 is the only one most people need.”
We checked all seven against the GitHub API on Aug 17 and the post's only inaccuracy is modesty: every count is in the right ballpark or undersold. markitdown is actually 174,189 stars (claimed 169k), ponytail is 104,530 (claimed 90k — 14k under), awesome-mcp-servers 92,474, crewAI 57,202, goose 52,902, skyvern 22,769, mcp-use 10,494. The one-line descriptions match what each repo actually does, and — credit where due — the video described ponytail only as “makes AI write LESS code” and never cited its famous 80–94% number, which the maintainer's own README has since walked back to ~54% less code in agentic use. The only stretch is “pulled from GitHub today”: the counts read a few weeks stale. “Number 1 is the only one most people need” is opinion, but a defensible one.
What holds up
- All 7 repos exist; every star count is accurate or UNDERsold vs live API numbers (markitdown 174,189, ponytail 104,530, awesome-mcp-servers 92,474, crewAI 57,202, goose 52,902, skyvern 22,769, mcp-use 10,494 — Aug 17)
- One-line descriptions match what each repo actually does
- The video did NOT repeat ponytail's retracted 80–94% less-code claim — restraint most listicles skip
What doesn't
- “Pulled from GitHub today” reads a few weeks stale — ponytail is 14k stars under its real count
- Ponytail context the slide omits: the maintainer's own README corrected the viral 80–94% single-shot benchmark to ~54% less code (agentic), calling the old gap “partly a conversational-baseline artifact”
- “Number 1 is the only one most people need” is a take, not a fact — though markitdown is genuinely the broadest-utility pick
The catch
Even the honest lists have a shelf date — “pulled today” is the one claim in this post that didn't survive checking, and it's the one that cost nothing to get right.
How to actually do it
- Start with markitdown if you handle documents at all — any file to clean markdown is the broadest-utility tool on the list
- Before adopting ponytail, read its README's corrected benchmarks (~54% less code, ~20% cheaper agentic) — the viral 80–94% number is retracted
- Verify any star count yourself in ten seconds: api.github.com/repos/OWNER/REPO, look at stargazers_count
Seven for seven — a listicle that undersold its own numbers, which almost never happens.
- Confidence
- High
- Posted by
- an AI-and-tech repo-listicle account — no gate, no course, just save-bait
We test hype for free. We build the real thing for a living.
Thirty minutes, no pitch — and you'll leave with something useful either way.
Book a call with Todd or start with the free Business Checkup →The Verdict Weekly
Three verdicts every Friday. Free forever, unsubscribe anytime, no spam — that would be ironic.
© Schreier Group · schreiergroup.com · See a wild AI claim? Drop it here and we'll test it.