← The AI Hype Audit — all 115 verdicts
PARTLY “6 open source repos I actually use every week” — five check out, one invents a benchmark
The claimSix GitHub repos worth your time: Hermes Agent (229k stars, “feels 10x faster than Claude Code”), Agent Browser (40.4k), Career-Ops (63.5k), Crawl4AI (77.8k), yt-dlp (184k), Headroom (65.9k — “20% fewer tokens for coding agents”).
Rare for the repo-listicle genre: the star counts are real. We checked every one against the GitHub API — Hermes Agent genuinely sits at ~231k stars, Agent Browser at ~40.7k, Career-Ops at ~63.9k, Crawl4AI at ~78k, yt-dlp at ~184.6k, and Headroom at ~66.4k (the slide's owner slug is a typo — it's headroomlabs-ai/headroom, not labs-ai). The descriptions match what the repos actually do. The one invention is the punchline: “feels 10x faster than Claude Code.” Nobody credible measures that — the maintainer's own claim is about 40% faster on tasks once the agent has built up 20+ skills, and independent 30-task head-to-heads land at rough parity, with Claude Code ahead on output quality and Hermes ahead on autonomy. 10x is a felt number doing benchmark cosplay.
What holds up
- All six repos exist and the star counts are accurate to within rounding (GitHub API, Aug 15 2026)
- Descriptions are faithful — Crawl4AI and yt-dlp especially are exactly what the slides say
- Career-Ops and Headroom are real breakout projects, not vaporware
What doesn't
- “Feels 10x faster than Claude Code” has no basis — the vendor's own number is ~40% on skill-heavy tasks, and independent tests show near-parity
- Headroom's owner slug is wrong on the slide (labs-ai/headroom 404s — real repo is headroomlabs-ai/headroom)
- Installing six agent frameworks because a slideshow said so is how you end up with a toolbox and no product
The catch
When every number in a listicle checks out, the one adjective that doesn't is doing all the selling. “10x faster” is the only claim on these slides you can't verify with one API call — and it's the one that makes you tap follow.
How to actually do it
- Verify any star-count claim yourself in ten seconds: api.github.com/repos/OWNER/NAME — look at stargazers_count
- Pick ONE tool for a real job you have this week. Crawl4AI if you need site data an agent can read; yt-dlp if you archive your own source footage
- Before adopting an agent framework, run one real task from your actual business through it — 30 minutes tells you more than any star count
- Confidence
- High
- Posted by
- an AI-news slideshow account (vacation-photo background format, follow-for-more funnel)
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.