The AI Agent Cohort Program is a flagship course Raymond Hou launched in August 2026, subtitled “Super AI Individual” (see The Super AI Individual: Extending a Working System from People to AI). Its official interactive quiz and curriculum entry point is ai.lifehacker.tw. It works alongside the AI Agent Mini Course, released in the first half of 2026: the mini-course is for outreach, introducing more people to how they can start building an AI double; the cohort program is for implementation, helping students equip themselves with methods and connect them into a sustainable AI team, supported by a year-long coaching community, live co-learning, and a platform for student projects. Raymond sums up the two courses this way: “The Super Individual’s Way of Working is a work system for people today; Super AI Individual is a work system that lets AI do tasks for you and amplify your productivity.”
Origins: From Mini-Course to Cohort Program
In the first half of 2026, Raymond had originally scheduled a period of rest, but “broke his own promise” and launched the AI Agent Mini Course. Agentic AI had made him feel strongly that “this is AI in its complete form,” and he was eager to share the world he saw and ways to avoid unnecessary detours. After launch, the mini-course accumulated more than 100 positive reviews. Several students called its format—primarily visual and text-based, packaging experience into Skills that could be directly given to students’ AI—“the most unusual course format I’ve ever taken.” Many then asked for a student-only community and a boot-camp-like coaching program. Those requests became the starting point for the flagship course.
The public record of course preparation also became part of the content. Starting in July, Raymond published a series titled “Posting Every Day Until the AI Agent Cohort Program Launches,” sharing his preparation process. By late July, the course content had been updated more than 364 times, serving three kinds of students: those with no experience, those who use AI but feel confused, and those who already have some experience.
Key Decision: Partnering With ShiFu to Distribute the Course
This was the first time Raymond distributed a main course through a platform outside his own channels. He unusually wrote a long letter to former students and readers explaining the difficult decision, titled “Why I’m Giving My Course to Another Platform to Sell—and Raising the Price.” The main reason matched his logic for using AI Agents: delegate work that he is not good at and that drains his energy:
- Raymond sees his strength as “making the content good and ensuring the delivered product meets a quality standard I believe in.” Marketing is “the most mentally exhausting part,” while customer service took too much of Yuzu’s time. Partnering with the platform allows a professional team to handle reach while he focuses on content and interaction with students.
- On pricing, he was direct: the revenue share with the platform is one cost, but the deeper reason is the delivery format: “If I priced it at a few thousand dollars, I wouldn’t be able to stay with you for a year. That’s the truth.”
- At the same time, he deliberately kept free and low-cost entry points: free public Claude Code resources, the low-cost mini-course, and a newsletter that remains free. He explicitly said, “You don’t need to buy this course to get started,” and “The last thing I want is someone forcing themselves to swipe a card because they’re afraid of falling behind.”
This decision follows his long-term positions in Trust Over Traffic and The Product Funnel: From Mass Traffic to a Deep Community of One Hundred: free content builds trust, paid products provide deeper delivery, and marketing to “make the pie bigger” is delegated to those more skilled at it.
Launch Timeline (2026-07 to 09)
- July 28 and 30: Two free trial livestreams for former students. Shortly after the first began, the server crashed under the incoming traffic. Raymond was “completely stunned, and immediately furious” on stage. A student’s comment in the community—“The incident actually proves it wasn’t prerecorded! I just signed up and can’t wait for the full course!”—turned the evening around. That same day, Raymond finalized his answer to a prompt on an interview board: “Can’t outrun them? Then run toward the sea.”
- August 1 at noon: Course materials and the student community opened on time for former students who had purchased access. The onboarding pipeline (purchase verification, GitHub, Discord, and the student project platform) handled 237 applications that day. Sixteen people got stuck in the invitation process for a time; the issue was fixed that day. Raymond wrote in his diary: “Today I actually felt very calm, because I know software and systems evolve by colliding and correcting.”
- Early September: All main-course content went live. Raymond published a long reflection explaining why this might be his “last tool course”: as AI capabilities overturn how individuals learn and meet more than 80 percent of tool-teaching and customization needs, traditional courses teaching the operation of a single tool have lost value. The cohort course instead focuses on building one’s own digital-employee doubles and irreplaceable practical workflows. 1
An Experiment in Course Format
The flagship course extends and expands the mini-course’s format experiment: monthly live coaching replaces prerecorded courses that are filmed six months in advance, purchased by students, and left unwatched; Raymond’s practical experience is packaged as Skills and given directly to students’ AI. The course also includes a deliberately designed AI teaching assistant, “Raymond Hou’s AI Double That Won’t Give Answers” (see Raymond’s AI Double). Its persona is “help you ask the right question, not solve the question for you.” Students tested it with various trick questions on launch night, but it held its boundaries and earned the nickname “Socrates-mong.”
Position in Raymond’s Framework
This course is the product implementation of The Super AI Individual: Extending a Working System from People to AI and a sequel to The Super Individual’s Way of Working (55-Unit Online Course): the former teaches a work system for “people,” while the latter teaches a work system for “AI.” The decision to “hand marketing to the platform” is itself a demonstration of the course’s central claim: delegate work that is not worth doing by hand, and do so with confidence (see From Hours Worked to Outcomes: Performance in the AI Era).
Source
Footnotes
-
Raymond Hou, “This May Be My Last Tool Course: Why Build an AI Agent Cohort Program Priced Above NT$10,000?” public Facebook post, 2026-09-04. View original ↩