✦ Common themes from student feedback
- The Dragon Ball / Summon Heroes metaphor is the most memorable and actionable — multiple students still use it weeks later.
- The homework guides are the real value — progress happens in the hands-on work, not just listening.
- Students already using ChatGPT / Gemini / Copilot say this course teaches "fundamentals" and "human-centered AI", not another set of tips.
- Real challenges exist: non-technical students find some terminology unclear; advanced concepts need more examples and time.
— Real, unedited student feedback.
★ Featured S3 · Freedom Non-tech 2026-04-21
❤️ Using the Method When It Mattered Most
"You feel more prepared, more well-rounded, and more at ease — you no longer feel totally helpless."
After the first three sessions, he used "three-layer shuttle + hero perspective" to cross-validate a treatment plan for a family member — summoning three real, world-class experts to examine the same plan from their own angles: 2018 Nobel laureate Tasuku Honjo, Suzanne Topalian of Johns Hopkins, and Julie Brahmer, who specializes in the relevant clinical trials. The course did not replace professional medicine — he still takes clinical questions to doctors. But facing those doctors, he felt more grounded, more prepared, more at ease. This is what People's AI actually looks like: not replacing expertise, but making an ordinary person less helpless in the highest-stakes moment of their life.
— Jack · Chinese Cohort 1 student (used the method to cross-validate a family member's treatment plan)
★ Featured S4 · Vision Engineer 2026-04-24
⚙️ A 30-Year Chip Engineer Turns Experience Into an Agent
"The further I get, the more I feel none of the earlier homework was wasted."
A silicon-photonics chip design engineer at Broadcom with 30+ years of experience and zero formal AI-engineering background before the course — he learned with Claude Code alone. He independently built a CPO (co-packaged optics) analysis agent, packing three decades of design experience into a three-layer structure — knowledge base + rules + execution harness — and validated it against three real published papers. On the same set of parameters, accuracy jumped from a 67.5% baseline to 100%. From first contact to a verifiable shipped artifact took just 17 hours, working through the night. He is not an AI researcher, yet he had already turned his professional workflow into a genuinely usable agent. This is what "domain experts can build" looks like.
— Steve · Chinese Cohort 1 student · Silicon-photonics chip engineer at Broadcom (creator of a CPO analysis agent)
📈 Accuracy 67.5% → 100% · shipped in 17 hours
★ Featured S2 · Leverage Non-tech 2026-04-23
🧭 One Homework Assignment Overturned Years of Startup Fantasies
"I've wanted to start a company for years. I never expected that in one day it would overturn all my old ideas — and then I quickly found a new direction."
He ran the S2 homework three times. First pass: he put several long-held startup ideas before the "hero panel" — all rejected. He was grateful: "It saved me a fantasy I'd carried for years." Second pass: he rethought, still too generic. Third pass: he summoned seven startup mentors from across the US and Canada, sat them down together, funneled a pile of ideas down to three, had the mentors analyze each pain point, and locked onto a new direction. One homework assignment overturned years of startup fantasies and quickly surfaced a way forward. He resubmitted the work at 1:30 AM.
— Chinese Cohort 2 student (used the method to find a new career direction during a period of unemployment)
★ Featured S4 · Vision Engineer 2026-05-02
🧬 A 30-Year Medicinal Chemist's Peptide-Drug Review Agent
"It demonstrates deep molecular-mechanism analysis that ChatGPT cannot do."
A medicinal chemist of thirty years who once led the building of a drug-discovery AI platform. He built a cyclic-peptide drug-discovery project review agent, grounded in the real peptide-chemistry authority Victor J. Hruby, analyzing layer by layer from molecular recognition to conformational control to receptor selectivity. His S4 homework scored 95, production-grade. This proves one thing: a domain expert + AI = a domain Wukong. Thirty years of expertise is a moat ChatGPT cannot learn — but an agent can carry it.
— Workshop student · medicinal chemist of 30 years (creator of a peptide-drug review agent)
★ Featured S3 · Freedom Engineer 2026-07-07
💬 Workplace-Communication AI · My "Training Partner"
"AI doesn't replace thinking — it helps us think deeper and work faster."
The future of AI is not a zero-sum "human vs. machine" showdown — it is a deep and sweeping "human-machine recombination." All the system design in my day job is done with the help of multiple AI tools: AI gathers, reads, and summarizes material fast, while I keep cross-examining it — probing, challenging, iterating — to refine the design. Thanks to the previous AI course, I also built a workplace-communication AI that became my "training partner." On my commute I rehearsed aloud with it, polishing the presentations for two complex-system ADRs — and won leadership's approval.
— Joy G. · Chinese Cohort 1 alumna · Systems engineer (creator of a workplace-communication AI)
🎤 Two complex-system ADRs approved by leadership
★ Featured S3.5 · Fast-Forward Non-tech 2026-05-07
🌟 Exemplary Work · A User's Tears of Joy
"My mentee cried after testing it, and recorded a video to share how it felt."
She sent her first Compass Agent to a book club mentee to test. The mentee had been using ChatGPT for a long time. She deliberately used a fresh account to avoid bias from prior chat history. She said this conversation felt completely different from ChatGPT — she felt understood and seen.
— Sufeng, S3.5 Workshop Student (Navigation Framework / Compass Agent creator)
★ Featured S3 · Freedom Non-tech 2026-05-04
🌟 Top AI Score · Government Benefits Guide Agent
"The user's ignorance is not a flaw — it's the reason you exist."
This agent doesn't serve "low-income people" as an abstract category — it serves the Mexican-American single mom working two jobs in Texas who opens it at 11pm after her kids sleep, the housekeeper in California who doesn't know "Medi-Cal" and "medical assistance" are the same thing, the just-laid-off delivery driver in Georgia who's always worked and never thought he "qualified." Five iron laws — the hardest one: never let a user leave a conversation empty-handed. Always deliver a concrete program + a clickable link/phone + one line "let's start here." At the language layer it's a translator — the user says "I can't pay my electric bill," the agent internally translates to LIHEAP; to the user it never says acronyms, it says "electric bill subsidy." User ignorance is not a flaw — that's exactly the reason it exists.
— S3 Student
★ Featured S2 · Leverage Engineer 2026-04-23
Conversations Across Time
"After adopting the 'summon a hero' prompting technique, I felt the same joy Liu Bei must have felt 1,800 years ago — a sense of having conversations across time with the great masters of human civilization."
— Upwork Freelance Software Engineer, S2 Student
★ Featured S1 · Courage Engineer 2026-04-04
AI Flight Agent
"By giving AI the role of a flight booking agent, it gave far better answers — saving $300-400 per ticket. Four tickets, over $1,000 saved."
— Sunny, S1 Student
💵 Actually saved $1,000+
S2 · Leverage Non-tech 2026-05-08
It turns out that when I'm hit by sudden pressure in life, AI can help me calm down — no more panic, no more feeling lost.
I had fallen into massive anxiety from a sudden life change, with no idea where to begin. After re-watching the class materials, I used the summon-a-hero method to build a "Professional Advisor" hero and asked it to help me analyze how to respond. The kind of consultation that would have been extremely expensive with a real professional — I got a full answer at a fraction of the cost. But what surprised me more was this: it gave me the feeling of being seen by reason itself, and both the panic and the sense of being lost were soothed. The AI era really is here. Thank you, teacher, for opening this door for me.
— S2 Student · Self-help professional consultation scenario
Other Non-tech 2026-05-07
I've lived nearly 50 years, and this is the first time I've gotten addicted to doing homework.
It's because the teacher designed the assignments with real care — they build up in layers, solving real problems from angles I could never have thought of myself. Every AI course I'd taken before was all about the technology, never truly people-centered. Yours is the most grounded, the one that actually gets to the root.
— Cross-cohort student · 50+ years old
S4 · Vision Creator 2026-07-29
"The core product is not AI-generated stories — it is the preservation of authorship."
He built a storytelling companion agent for pediatric cancer patients — the child supplies plot and characters, the AI organizes and polishes, but never overwrites the child's own authorial voice. The expert persona is grounded in the real pediatric hematology-oncology and narrative-medicine specialist Trisha K. Paul. His English-cohort homework climbed steadily, reaching 92 at ENS4. His product thinking is mature: he plans to validate with 5-10 patients using a "human-powered agent" before writing any code. In his words: "Generic AI is excellent at polished prose. That is precisely the danger — a polished story that could have been written by any child is a failure."
— — English-cohort student (creator of a storytelling agent for pediatric cancer patients)
S4 · Vision Engineer 2026-05-27
"The simplest MVP is to validate commercial viability — if it doesn't hold up, we immediately try another idea."
An enterprise AI-deployment consultant in Los Angeles. He deploys private LLMs on on-prem private servers for enterprises, has built a token-aggregation framework, and does local small-model deployments for privacy-sensitive verticals (counseling, legal, healthcare). He has already commercialized exactly the "enterprise privacy opt-out" direction the platform is planning, and identified a concrete LA Chinese-SMB vertical (home purchase → renovation → loans), adopting an "intake agent + human escalation" monetization model.
— — Workshop student · enterprise AI-deployment consultant
S5 · Refinement Engineer 2026-05-31
"I want to invest in myself. I could just coast, but while I still have the ability and the ideas, I'd rather give it a shot."
An Intel semiconductor engineer, around 64, Tsinghua EE class of 1982, who had already used Claude Code for about a year before joining. He built several agents for his own use — a semiconductor-industry briefing agent (run monthly/quarterly, 36KB+ per report), a car-search agent, a flight-booking agent — and planned daily-news and chip-status agents. Real production-grade personal use; he hit his daily quota in a single day. He proactively brought the teacher a "human-in-the-loop paid service" business model: "Companies often ask me to share industry views or join meetings for pay, and I always decline. But if it's AI-generated and human-reviewed to remove errors, I think that's workable."
— — Workshop student · semiconductor engineer (built multiple agents for his own use)
S5 · Refinement Non-tech 2026-05-31
"Stop. Whether your Tech Lead is a jerk is next week's politics problem."
She built True Self Shield — a mental-recovery system for high-pressure knowledge workers, designed for Silicon Valley tech workers (interview freeze, harsh code reviews, project stress), with a "mental legion" of four archetypal mentors and a "forced-interaction seal" that cuts off spiraling rumination and won't let the user proceed until they confirm. She broke through to 92 at S5. The teacher called her forced-interaction mechanism "a genuinely deep cut" that "severs the long spiral like a cold law of physics."
— — Workshop student (creator of a mental-recovery system for high-pressure knowledge workers)
S6 · Mastery Educator 2026-05-25
"I'm in — let's build this together. From now on I'll spend more time on AI education."
A former university teacher (China) who moved into US IT, based in Atlanta, with 5-6 years volunteering to teach high-schoolers to code. She built TechBarRaiser — a tech-role interview coach whose feedback style leads with "name the gap first." It's published on the platform, rated 5.0 stars, free tier, ~7 uses per month. Perfect 6/6 attendance across the cohort (only 17 of 396 students hit 6+ homework), scores rising monotonically 75 to 92, clearing the Demo Day publish bar. This is the cleanest proof of a student shipping a live, usable public agent.
— — Chinese Cohort 1 student · former university teacher (creator of the live agent "TechBarRaiser", rated 5.0 stars)
S3 · Freedom Engineer 2026-05-25
"You gave me a solution in one minute — something I'd been stuck on for ages."
A data scientist. She built a two-person de-escalation agent — it analyzes the real need underneath an argument and reframes it for the other side. She'd already prototyped it with friends; her S3 homework broke through to 93 and friends' feedback was strong. The product form factor is still being refined, but the direction is clear: turn a one-minute de-escalation into a tool you can carry.
— — Workshop student · data scientist (creator of a couples de-escalation agent)
S2 · Leverage Other 2026-04-24
"This is exactly the help I needed — thoughtful, detailed, cutting straight to the essence. I've already recommended your course to several people."
A course student who, busy with other things, only started homework at S2. His very first message was unprompted gratitude with no ask attached — and he had already recommended the course to several people and signed up for the paid track. A pure word-of-mouth + referral + paid signal.
— — Chinese-course student
S3 · Freedom Non-tech 2026-05-31
"Putting a data scientist's rigor and a life coach's warmth into the same agent."
A staff-level data scientist who is also a CTI-certified life coach, bilingual, focused on faith and whole-person wellness. She's working on several agent directions including a faith life-coach, broke through to 92 on both S2 and S3, and confirmed for Demo Day. She represents a distinctive direction signal: a data scientist's rigor and a coach's warmth in one student.
— — Workshop student · data scientist + certified life coach