Summary

AI is compressing both entry-level jobs and the barriers to entrepreneurship; the real dividing line is whether you can turn it into personal leverage. Gary Tan shared a figure at YC Startup School: in the US, more than 6% of computer science graduates are unemployed after graduation, versus about 3% of arts graduates. But Chenran has also seen AI let newcomers cross disciplinary barriers and independently complete an MVP—“the leverage of one person has gone up.” The episode presents the simultaneous contraction of the employment funnel and emergence of the super-individual.
The early upside of a new paradigm belongs not only to people with technical backgrounds, but to those without legacy baggage who learn and validate faster. Chenran calls herself “a purely AI-native worker.” The roughly three years from ChatGPT’s launch in November 2022 to context engineer becoming a recognized capability standard were, in her words, “a period of wild growth”; she filled 10 pages with AI ideas, some of which later became commercial products or projects she is now building.
Sustainable irreplaceability, several guests suggest, rests on 2 dimensions: working with people and continuously mastering new tools. Leo sees AI as an “enabler” that will enter every industry, “regardless of human will”; Gao Shutong worries that AI could replace the foundational work done by junior business analysts and computer science graduates. Ryan believes investing, art, the humanities, and taste are difficult to quantify, while simply “racing AI” is dangerous because individuals have little way to know whether their exponential growth can keep them ahead.
Building in public is evolving from personal expression into compounding career infrastructure. Chenran sees it as a mechanism for slower thinking, a verifiable social calling card, and a trust asset: her content, demos, and public interactions continuously sell her for her. “I got my current job because of building in public.”
The 4 young people share a strategy of shortening the feedback loop on self-knowledge through real experiments, rather than getting one decision right at the outset. Leo scored professors, VCs, quantitative researchers, patent lawyers, and management consulting one by one; Shutong tried research, dentistry, and medical school before abandoning a path she had pursued for 5 years; Ryan gave up an overseas offer to found Spark Lab. Leo’s formulation is: “Try a round of things first,” then ask whether the worst-case consequences are bearable.
AI could recast companies from organizations that must support 10 or 100 people into small businesses that need only support their creators. Ryan therefore calls AI “a Renaissance, not an Industrial Revolution”: super-individuals and “one-person unicorns” prompt young people to ask earlier who they are, while making it possible to turn genuine interests into sustainable business models. Spark Lab’s methodology is “Find your spark by getting lost,” accepting that confusion and discovery are 2 sides of the same coin.
Hiring signals are also shifting from static metrics toward underlying drive and evidence of action. Bonnie says she will not look only at grades or internships; she asks what kind of person someone is, where their goals came from, what they have done to pursue them, and how they have grown. Making a job search more legible still starts with answering “Who are you, and what do you want?”—not with finding a standard answer.

Deep dive

1. Career entry points across 3 generations have shifted from clear paths in boom times to proactive choices under high uncertainty

The episode opens with a counterintuitive figure: Gary Tan shared at YC Startup School that more than 6% of US computer science graduates are unemployed immediately after graduation, versus about 3% of arts graduates. “The share of computer science graduates who are unemployed is even twice that of arts graduates”—the old ranking of major-by-major returns is losing its force.
Bonnie recalls that when people born between 1990 and 1995 graduated, finance, internet companies, major tech firms, brokerages, investment banks, and consulting were still obvious paths. Exchange programs, overseas graduate study, and campus entrepreneurship were also popular. IDG began focusing on the post-90s generation in 2013–2014 and was tracking and investing in several projects at the time; the episode mentions Liu Jingkang, whose project went public recently, as well as Bilibili, Lilith Games, and Heytea, though the original wording around the project name is unclear. That wave of investment made peers think, “Maybe we have a chance too.”
Ronghui looks back at graduating from college in 2008 and graduate school in 2011. Multinationals were still highly regarded, while mobile-internet startups offered opportunities to build from zero to one everywhere. At the time, he thought “everything was normal”; only later did he realize it was “a super-boom period of rapid economic growth.”
The host’s generational conclusion does not treat the present as uniquely difficult: Bai Yansong once told a student complaining about employment, housing prices, and eldercare burdens, “This is a question about youth. Every generation has its own problems.” What does not change is that there are always young people; what changes are the problems they face.

2. Chenran used content creation to navigate a crowded computer science market, while building a cross-disciplinary skill stack

Born in 1999, Chenran studied computer science at Fudan University and the intersection of computer science and entrepreneurship at Cornell. She now leads AI video products at OneTwoX. She describes herself as “a fairly integrated person”: programmer, product manager, director, social-media operator, and content creator all at once.
She started creating content through a microfilm elective in her freshman year, later making videos for campus events. When the vlog boom arrived in her junior year, she spent about 4 hours a week editing and kept publishing weekly, both practicing on-camera communication and preserving her college memories. “How do you spend your time in a high-quality way?” was the original motivation for this side track.
She did not branch out because computer science was unpopular; quite the opposite—it had become too popular. Talent was oversupplied, roles were heavily segmented, and once workflows were standardized, experienced hires naturally became more valuable in the market than new graduates. At the time, she found hot topics such as facial recognition and natural language processing “all very boring,” while entrepreneurship felt like “a fantasy.”
Chenran’s view is that growth can mask problems in processes and work quality. Once growth ends and an industry hardens, fine-grained specialization turns into intensified competition, leaving young people feeling that work is “increasingly meaningless.”

3. Once career paths were forced to diversify, peers lost a single scale for comparison

Few people around Chenran actually joined major tech companies: some became teachers, some pursued PhDs, some stayed overseas, and others joined major tech firms. “Everyone accepted the situation and found their own way to make peace with it.” Destinations varied wildly, but were not as bleak as imagined.
The cost of diversification is higher psychological pressure; the benefit is less comparison: “You can’t compare,” because everyone’s path is no longer measured by the same yardstick. It may not have come from ideal choices made proactively, but it still expanded the range of lives that could count as normal.
This explains why the episode keeps returning to exploration. Once standard answers stop working, young people cannot define themselves solely through a major or first job; they have to use projects, internships, public creation, and real collaboration to determine which capabilities can form a combination advantage of their own.

4. After dissecting the professor track, Leo rewrote “ideal” as a probability that updates with evidence

Born in 1997, Leo studied at Tianjin University and earned a PhD at Cornell; he is now a consultant at McKinsey. In his third year of college, he strongly wanted to become a university professor. By his third year of graduate school, he realized that being a professor involved more than mentoring students, writing papers, and running projects: there was also grant writing, partnership development, tenure review, and public service. His energy would be fragmented, and many of these matters were outside his control.
He did not reject academia overnight; he describes the shift as a declining probability. It may initially have been 50%–60%, then fell to 20%–30%. A researcher role at a national laboratory was once an alternative, but he found that he enjoyed interacting with people and solving “up-to-date” problems, while long-cycle research could make topics and solutions lose their timeliness.
Around 2021–2022, climate change emerged as a new direction and funding began to flow in, so he tried a related VC internship. The 2 experiences showed him that differences among funds in time horizon, sector, style, growth, and vintage can themselves produce entirely different career experiences.

5. Career choices can be tested, but no spreadsheet can bear the consequences for you

Leo once made a list comparing 5 paths: university professor, venture capital, quantitative researcher, patent lawyer, and management consulting. His criteria included personal preference, career development, long-term planning, and whether he would continue to enjoy the work. He tried to experience each path firsthand where possible, and otherwise held in-depth conversations with practitioners.
His candid formulation was: “You can’t say I chose this job. I feel like this job ultimately gave me an offer.” Management consulting chose him, while he also gave it a chance during the application process. Exploration can improve the odds of a fit, but it cannot erase external agency or chance.
Leo’s decision sequence is to do first and feel it out afterward: continue if he likes it, reconsider if it does not fit, and diversify his efforts across options when a pivot becomes necessary. He now adds one more question: if the consequences deteriorate, can he bear them? That makes “try a round of things first” a risk-aware strategy rather than a dismissal of risk.
He refuses to impose a fixed definition on success: “I never define success.” If pressed, it means that he, his family, and the people around him are happy. Because his emotions easily affect others, maintaining a stable inner core is also part of success.

6. Shutong tried the 3 broad paths in medicine before finding the courage to abandon a direction she had pursued for 5 years

Born in 2000, Shutong studied pharmaceutical sciences and immunology at UBC and will attend Yale for an MPH. She saw roughly 3 paths from the major: medical school, dental school, and research—and tried all 3 in practice. She worked in a COVID-19 drug laboratory for more than a year, shadowed at a dental clinic, took the medical-school entrance exam, and then shadowed in a hospital.
Interest in the subject was not the problem; she “doesn’t regret the choice at all.” Only after entering the environments did she realize she was afraid of pathogens. After just 2 and a half weeks in the hospital, she became seriously ill and remained sick for a month. Her physical reaction made the path more than an abstract plan.
After eliminating the 3 routes, she spent a period unsure what else she could do with the degree. She used LinkedIn to find roughly 30–40 people with similar backgrounds who had later moved into VC, PE, investment banking, or consulting, including people who had switched into business after completing 4 years of medical school. The career map was not handed to her by a university; she assembled it through one conversation at a time.
The 2 and a half weeks in the hospital had already given her the answer. The hard part was finding the courage to “give up a path I had persisted in for 5 years.” Her later conclusion was to listen more closely to her inner voice: “You already have the answer in your heart. The question is whether you have the courage to trust it.”

7. The truly expensive mistake is not moving slowly, but being pushed by the peer clock into an unchosen path

Shutong does care about judgment, but she is “very selective about whose opinion I actually care about.” Most classmates in her major were still pursuing medical school, dental school, or research; as far as she knew, she was the only one to make this kind of pivot. Support and opposition would exist either way.
She has little instinct for numbers, and from envying competition gold medalists while working hard just to pass, she arrived at one conclusion: “You cannot become anyone else; you can only become yourself.” Looking back at age 70 or 80, if she had tried everything she wanted to try, regardless of the outcome, “it would have been an interesting life.”
More salient than AI is age and peer pressure. Some people panic at 25 about marriage, a PhD, or falling a step behind; they see someone get into a PhD program and apply themselves. But real choices require time and the cost of trial and error. What she fears regretting most is not failure, but “that you never tried.”

8. Ryan gave up an overseas offer to found Spark Lab, betting on the growth embedded in uncertainty

Ryan entered Zhejiang University’s information security program in 2020, began entrepreneurship in 2022, and launched Spark Lab in 2024 before graduating. The project encourages young founders to build from the inside out, then construct the world they want to create, while experimenting with an accelerator model not yet seen in China’s venture ecosystem.
The project initially ran mostly on intuition. The team could not immediately explain its value proposition or business model, but started anyway. Ryan also gave up an offer from a top overseas university and his status as a new graduate. He did not think much about it at the time; only in retrospect did he realize it was an important choice.
He kept investing in the project for 2 reasons: the work was interesting and meaningful, and there was ample room for career growth, influence, perspective, and network. “No one on the team started out for the money.” Compensation only needed to support them; they were more “obsessed with pursuing more uncertainty.”
Ryan believes the economic pressure and appetite for large purchases among people born after 2000 or 2005 are relatively limited. Buying a home in Beijing or Shanghai is too difficult, which in turn weakens the desire to buy one. Salary is therefore not necessarily the top priority; young people care more about a distinctive perspective, room to grow, platform influence, and interest.

9. AI’s impact is highly stratified; the shock experienced by frontier users does not represent most young people in China

The hosts initially expected the guests to be more anxious, but heard a more measured attitude. Ryan says, “Who wasn’t anxious in the first 3 months after ChatGPT appeared?” But 2 or 3 years later, people were better able to judge what stage AI was in and what role they could play.
He also cautions that “China is really huge.” Many peers do not even have computers; many may have only started using AI in 2025 because of DeepSeek, primarily through Doubao and DeepSeek. For them, it is merely “another thing similar to Baidu.” Ryan believes the lives of perhaps 90% of ordinary people in China have not changed noticeably.
Bonnie believes AI’s impact on the guests has been smaller than expected and will create more opportunities over the long term. Koji says AI is restructuring technology, tools, and efficiency, breaking boundaries and amplifying imagination, but has not changed foundational needs such as emotional resonance, self-actualization, a sense of value, curiosity, and the desire to grow. Ronghui adds that AI has also unlocked capabilities for people who are not good at organizing collaborators.

10. The most effective antidote to anxiety is not prediction, but an action buffer that allows failure

One person in the research took a year off from school simply to think about what they wanted and where they stood. Coy, a junior on the Spark Lab team, also took a leave of absence to experience different jobs before deciding on a university and career direction. The host was struck by the maturity Coy displayed in collaboration, far beyond the stereotype of “a college junior.”
Several hosts summarized the common trait in these stories as “vigorous curiosity.” More information and fiercer competition create anxiety, but finding your own path starts with taking the first step. “Try something here, poke at something there, look around in every direction”—action is the antidote that shortens the feedback loop on anxiety.
Koji relayed Paul Graham’s 3 conditions for great work: being good at it, loving it, and finding it “effortless.” The third means having an edge in doing the same thing, receiving positive feedback more easily, and being more willing to keep investing. The broader the exploration, the greater the odds of finding that fit.
Bonnie describes IDG as a “buffer zone for trial and error.” The i-Star Program lets undergraduates, master’s students, and PhD students experience VC; the PhD Program and JD Program allow participants to treat investment firms as a “second laboratory.” After rotating through roles, some stay in investing, some start companies, and others return to industry or research.

11. Chenran sees the 3 years after ChatGPT as a standards-free period of wild growth for young people

When Chenran encountered ChatGPT in November 2022, she did not immediately form a grand thesis; curiosity came first. She had studied natural language processing and could also write, shoot, and communicate, so she quickly saw that her communication skills fit naturally with AI: “I had no baggage. I’m a purely AI-native worker.”
She observed that prompt engineer was later formalized as context engineer. It took the industry roughly 3 years to define a standard for this capability, and before that standard emerged was “the period when everyone was growing wildly.” Newcomers did not need years of experience; they only needed to be more willing to experiment and faster to understand.
AI also freed her from choosing a single identity among senior product manager, frontend, backend, and full stack. Industry knowledge that once appeared deep and specialized could be explained by AI from the basics upward. “The barriers in every industry have come down,” allowing her to conduct real experiments across multiple fields.
After ChatGPT launched, she filled 10 pages of draft paper with ideas, later discovering that some had become commercial products or projects she was working on. More importantly, AI can help validate ideas: an MVP that once required a team might now be achievable by one person.

12. Building in public compresses thinking, trust, and opportunity into one compounding loop

Chenran first treats building in public as a discipline of thought. Writing down a view publicly forces her to slow down and reorganize knowledge, much like the Feynman technique. Once published, the content leaves a record, invites comments, and attracts people who “point fingers.” After overcoming the initial embarrassment, she believes the benefits clearly outweigh the costs.
As content, demos, and interactions accumulate, they become a “detailed social calling card,” showing how she thinks and the network around her. Strangers can build trust faster; she no longer has to laboriously sell herself every time, and people may even approach her with information and opportunities.
The most direct result is: “I got my current job because of building in public.” Age has also made her clearer about what she does and does not want. Compared with financial freedom, she now wants “attention freedom”—the ability to decide where her attention goes.

13. Leo reduces personal moats to learning and interpersonal skills, while entry-level disruption could affect later promotion

Leo’s first-choice tool has shifted from Baidu to ChatGPT and related products. He distinguishes traditional machine learning, deep learning, and generative AI: older models primarily processed data, recognized patterns, and made predictions; after ChatGPT, they began directly generating knowledge and content.
In his view, AI is an “enabler” that will inevitably enter every field: “It does not depend on human will, it is uncontrollable, and it will definitely happen.” The question is no longer whether to use it, but who can master new tool paradigms faster and keep their knowledge evolving with those tools.
His formula for irreplaceability has only 2 components: dealing with real people, stakeholders, and the physical world, while continuously learning languages, programming, and new tools to solve problems. “At its core, it is still learning ability and the ability to deal with people.”
Gao Shutong worries that the basic work handled by junior business analysts at large consulting firms and by computer science graduates could be replaced by AI. If newcomers cannot obtain those entry-level jobs, they may later struggle to gain experience and move up the ladder.

14. Shutong treats AI as a team of employees, and reframes failure as training for identifying her strengths

Shutong knows she is not naturally strong with numbers, so she will not force herself into quantitative finance or computer science simply because those fields are popular. The more automated and competitive they become, the more likely only people with genuine aptitude and passion are to stand out. Her strategy is to identify “what will not be replaced” and strengthen the advantages she already has.
She treats Perplexity, Gemini, ChatGPT, Manus AI, and Dance Spark as “junior employees.” Some are good at summarizing and planning; others are good at making PPTs. The point is not to chase every tool, but to ask like a manager: “How can I make good use of my employees to work for me and achieve my goals?”
After giving up medical school, she was lost for at least 6 months. On her birthday, friends wished her “a sense of direction in life.” She later became grateful for the experience of supposed failure because it taught her how to choose and trust her inner voice. Faced with anxiety about AI, the only thing she can control remains “doing the current thing well.”

15. Ryan sees AI as more like a Renaissance: it reduces organizational scale and awakens individual agency earlier

When Spark Lab launched in 2024, its first observation was that AI tools had significantly lowered the barrier to getting things done, creating “super-individuals” and “one-person unicorns.” In the past, a company had to support 10 or 100 people; now one person may be able to make the work and only needs the business to support its creator.
This means interest and business model no longer inherently conflict. Young people can build companies around what they genuinely care about, creating business models sufficient to support themselves rather than treating employment inside an organization as the only answer.
Ryan’s second judgment is that “AI is a Renaissance, not an Industrial Revolution.” It makes people value agency again and strips away some of the glamour of major tech companies and “screw” jobs. In the past, many people only began asking what they liked in their 30s; now they can ask the question much earlier.
The price is earlier confusion, but confusion and agency are not mutually exclusive. Spark Lab’s slogan is “Find your spark by getting lost”—“you can find it only after getting lost.” Discovering passion sometimes requires allowing yourself to temporarily lose direction.

16. The last jobs to be replaced may be those involving people, while hiring shifts from metrics toward genuine drive

Ryan believes jobs directly involving people will be replaced later. Investing, art, and the humanities rely on elements such as aesthetics and taste that are difficult to quantify or transmit. This path is not about racing AI on speed, but choosing a value dimension that is relatively orthogonal to it.
Another path is to stay ahead of AI at all times, as researchers and people pushing the boundaries of human civilization do. But Ryan says this is “quite dangerous,” because it is difficult to estimate whether an individual’s exponential growth can continue to outpace AI. Like Go, it may simply be a matter of time.
Bonnie’s answer on hiring returns to the person. She does not look only at grades or internships, but at underlying drive, actions taken toward a goal, the thinking involved, and what the candidate learned and how they grew. A job search cannot be fully de-risked through a standard process; answer “Who are you, and what do you want?” first, and each step will acquire a clearer meaning.

Index Ventures Partner, Martin Mignot: Figma, Scale, Wiz: Inside Index’s Decacorn Factory Macro Talk 90: Aug–Dec: Stocks, Anti-Involution & Next DeepSeek Moment