About this data
SkillCurio maps 35 YouTube channels and websites for learning AI, coding, and automation. This page explains how listings are chosen and rated, where the numbers come from, how we make money, and how to flag something that looks wrong.
How resources are chosen
Listings are hand-picked, not submitted or paid for. We favor channels and sites that are actively publishing, have a track record of teaching AI/ML/automation skills, and cover one of four focus areas: hands-on coding, no-code automation, research & theory, or tool reviews & news.
No pay-to-rank, ever
No creator, course provider, or company can pay — in money, free access, or any other consideration — to be added to SkillCurio, to receive a higher skill-per-hour rating, or to move up in sort order. This is a firm policy, not a soft guideline.
Affiliate commissions (see "How we make money" below) are the only monetary relationship SkillCurio has with any listed resource, and they are structurally separated from ratings: the same person writing skill-per-hour justifications does not see or manage affiliate program applications, and a resource's affiliate status is stored as a separate flag that has no input into the rating logic. A resource can carry an affiliate link and a Low rating, or no affiliate link and a High rating — the two are unrelated by design.
What "skill-per-hour" means
Each resource gets a High / Medium / Low rating for how much practical, hands-on skill you're likely to gain per hour spent, relative to other resources in the same category — not a judgment of overall quality or production value. A long-form news roundup can be genuinely excellent at what it's trying to do (staying current on tools) while still rating "Medium" on this specific metric, because that's a different goal than a structured, hands-on tutorial.
The rating and its written justification (visible in each resource's detail card) are our editorial judgment based on watching/reviewing representative content, not a score submitted by the creator or generated automatically from engagement metrics. If a rating or justification looks off to you — including if you're the creator and think we mischaracterized your content — use the correction link below.
What goes into the judgment call. There's no hidden formula spitting out a precise number — the High/Medium/Low label is a reviewer's holistic call informed by the following, roughly in order of weight:
- Practical application — does the resource have you actually do something (write code, build a workflow, configure a tool), or mostly just explain concepts at you?
- Signal-to-filler ratio — how much of the runtime is direct instruction versus intros, sponsor reads, tangents, or repetition?
- Transferability — can what you learn be applied outside the exact example shown, or is it a one-off demo that doesn't generalize?
- Currency — is the content still accurate for how the tools/models work today, or does it describe an interface or capability that's since changed?
- Clarity — are difficult ideas actually explained, or assumed/glossed over in a way that would lose the resource's stated target skill level?
We deliberately don't publish a weighted numeric score (e.g. "73/100") because that level of precision would overstate the confidence of what is, honestly, a qualitative editorial read of representative content rather than a controlled measurement. High/Medium/Low is a call we're willing to defend and revise on request — a manufactured decimal score would not be.
Where the numbers come from
YouTube subscriber counts are refreshed automatically on the 1st of every month by a script that reads each channel's public subscriber count directly — no manual re-checking needed for that one number. Any channel whose subscriber count jumps or drops by more than 20% in a month is flagged for us to manually re-review (not auto-changed), since a sudden swing can be a signal that a rating or justification needs a second look.
Website traffic estimates (for the 10 non-YouTube resources, like course platforms and documentation sites), average video length, teaching style, skill level, and every editorial rating and justification are manually compiled and reviewed — there's no reliable public API for these, so they remain a snapshot rather than live data and will drift out of date between our manual passes. The dataset was last updated on —. Per-resource verification dates are listed in the table below.
The one exception on the live-data side is the "AI News" ticker on the homepage, which fetches live headlines from Google News on every page load and is not part of this static dataset.
How we make money
SkillCurio is free to use. To keep it running, a small number of listings include an affiliate link — currently the following:
Affiliate links only ever point to a platform's own official pricing or signup page — never to a
third-party reseller. Every affiliate link is labeled both on the resource card and inside its detail
view, and carries a rel="sponsored" attribute so search engines and browsers can identify
it as such. We are not currently running display ads or accepting sponsored/paid placements, though
that may change — this page will be updated first if it does.
Data by resource
When each listing's stats were last checked, and whether it carries an affiliate link.
| Resource | Type | Last verified | Affiliate |
|---|---|---|---|
| Loading… | |||
Changelog
Every meaningful change to SkillCurio — new learning paths, methodology updates, data refreshes, and site fixes — is logged publicly and dated. See the full changelog →
Found something wrong?
If a subscriber count, rating, category, or link is outdated or inaccurate — or you're a featured creator and want to flag a correction, update a link, or contest a rating — reach out and we'll fix it.