@maarcoofdezz Free Claude via Chinese models is when you save $20 but hand over all your code, data, and prompts to Chinese servers in return. Classic 2026: ‘I’m not paying Anthropic, I’m just feeding DeepSeek my IP.
Just another overused guide from the “faceless YouTube / AI channels” niche.
The guy posted 4 basic hook formulas (shocking number, “you think wrong”, in medias res, and direct call-out to the viewer) — the same shit that’s been chewed over a hundred times already on TikTok, YouTube, and Twitter.
Zero new ideas, just nicely designed pictures with examples that work about as well as any other clickbait formula.
The account itself promotes “passive channels without showing your face” → classic content marketing: “here’s free advice → later buy my course/proxies/service.”
In short — another empty nothing-burger for people who still believe that 4 template sentences in the first 10 seconds will turn them into a millionaire.
• They completely copied an already working format from channels like BeyondTheBlue (same titles like “We Can’t Explain”, red-and-white text on black backgrounds, high-contrast thumbnails).
• They generate content via AI very quickly (79 videos in 5 months).
• Short clickbait titles + emphasis on keywords.
At the end, of course, there’s a smooth transition into advertising their proxies: like, if you want to run multiple AI channels the same way at scale — take our proxies.
Classic scheme of “look how people make money on AI YouTube → here’s the tool so you can do it too.” Nothing new, just another hype around AI content farms.
The model names are real, but the pipeline itself - “matching via one model → execution via another” - is the author’s invention. Skills don’t work that way. Matching of descriptions and loading of SKILL.md happen inside a single session of a single model; there is no inter-model routing involved. The only real routing between Fable 5 and Opus 5 is the safeguards mechanism: some requests to Fable 5 are automatically sent to Opus 5, but that is about safety, not about “cheap matching + expensive execution.”
In other words, the author took real model names and invented an architecture around them that does not exist.