The young are useful to society as heretics, not as good citizens
Universities train young people to be good citizens, but what society actually needs is for them to question what everyone takes for granted; in the AI era every student can build their own projects, they just need to be allowed to.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
The young are useful as heretics
Amjad uses the sci-fi novel The City and the Stars as an analogy: in that city governed by an AGI, everyone is repeatedly reborn from compressed memories, until the machine breaks down and a genuinely new person is born, who immediately starts asking ‘what is outside the walls’. Amjad says the young are useful precisely for asking that kind of question, for being heretics. But today's universities do the opposite — they tell you ‘we have already figured everything out, just learn our way’, and train you into a good citizen: marry, have kids, buy a house, buy a car. The problem is that world no longer exists; you can perform brilliantly in college and still end up unemployed with a pile of debt.
— Amjad MasadSilicon Valley is itself a monoculture
Amjad says every society has blind spots, and Silicon Valley has plenty — it is a monoculture, with a fixed way of thinking, so many ideas simply will never grow out of Silicon Valley. Even the most radical culture is, at bottom, a monoculture. So you need young people to experiment, to attempt what is considered impossible, and occasionally someone breaks through: impossible yesterday, suddenly possible today because the technology changed. What universities should do is gather clear-minded, first-principles thinkers and encourage them to question the things that are deeply believed.
— Amjad MasadRaising money at 18 is an overused option
Gagan says young people today face a deeply unfair binary: either go to a four-year university and be governed by attendance, homework, exams and grades, or tomorrow someone hands you $50,000 to start a company. He thinks the investor path should be used far less, and only a very small number of teenagers should raise money. Most people at 17 or 18 know they want to enter the business world, but having two years to explore and learn without pressure is precious. He argues for a middle ground: two years to explore freely, with Silicon Valley actively sending information and resources your way, and more support if you have a good idea.
— Gagan VianiStarting too early is premature optimization
Amjad borrows a phrase from programming, ‘premature optimization is the root of all evil’: a lot of founding is premature optimization. He sees many kids now inside a circle where friends are raising big rounds and some have become unicorns, so the motive becomes a negative one like envy, and then they pre-commit to an idea they are not even interested in, optimizing in advance for a path they have not yet reached. He says the real question is what your curiosity is driving you toward; once you find the idea you are willing to spend ten years, a lifetime, on, it has enormous gravity and you cannot dodge it — he himself actively did not want to start a company, but the thing blew up into one on its own. He also says VCs now go to Stanford and give students a million dollars with no idea yet, so they have a cap table before they have even started, which is a bit like the music industry signing young artists.
— Amjad MasadFrom chess all the way to training models
Amjad says he has always swung between ‘wanting to build things’ and ‘wanting to read and explore widely’, the two feeding each other in a recursive process. He gives his recent obsession with chess as an example: he started playing very late, does not know why he was interested, then began writing programs to solve chess and training models, and found that LLMs are terrible at chess and hallucinate moves that do not exist. To figure it out, he — who had never trained a model — spent a lot of time studying how to train one. He realized from this that AI can now do machine learning itself, just as in 2024 Replit's agent suddenly learned to write code and afterward coding changed completely; machine learning may be at that same moment — he calls it vibe research or auto research. He is now building products at Replit to let anyone become a machine learning researcher, or let a boss with data fine-tune a model to solve a specific problem.
— Amjad MasadCompanies will become self-driving organizations
Amjad thinks the shape of companies will change. Silicon Valley has tried many experiments to remake companies — Holacracy, Medium tried it, GitHub tried a no-manager system, all failed spectacularly — but he thinks that was a technology problem: now you can let agents act as the connective tissue of an organization, turn bureaucratic processes into invisible things machines do in the background, give people real autonomy to work on what interests them most, and have AI judge your impact. He says he has already changed how management works at Replit, calls it a ‘self-driving company’, and that ultimately the company will run itself. He admits he does not have a fully formed idea, but is quite sure everything will change, and much faster than you think; if you do not natively understand this stuff, the transition will be hard.
— Amjad MasadTrust makes money land six weeks faster
Gagan says he volunteers in local politics in his spare time, and the biggest takeaway is understanding how valuable trust is. He gives a concrete comparison: in a nonprofit, after a donor verbally commits $50,000 to you, the probability that the money actually lands in your account is only about 20% to 50%; in Silicon Valley, that probability is about 75% to 100%. The difference is that if you are certain the money is coming, you can confidently move forward, and by the time the check actually arrives you have already advanced six weeks, which makes the check more valuable still. Conversely, without trust you change your approach out of fear the donor will be unhappy, drift from your mission, and are left with only anxiety and stress. He thinks this can be transplanted into schools: trust students to do what they want to do.
— Gagan VianiWith no grades there is no fulcrum to push against
Asked how to handle disruptive students — like Zuckerberg, who nearly got expelled for FaceMash — Gagan says there must of course be red lines, rules for things like sexual harassment and publicly humiliating others, and you need to build trust and a positive-sum atmosphere. But he says one core principle of the academy sidesteps the problem entirely: with no grades and no exams, the resistance students have to push against is already far smaller. So if you want to build FaceMash, go build it, and he is willing to defend that right even if he personally finds the idea weird and does not like it. He thinks that as long as you create an environment of trust and keep students from developing an anti-authority mindset, you unlock a huge amount of creativity and opportunity. Amjad adds that this drive to test boundaries can be channeled in healthy directions: kids who learn to program and hack on Roblox, rather than being sent to prison, could be put in a sandbox, have their talent discovered, and be steered into cybersecurity jobs or even government.
— Gagan VianiIn their own words · checked verbatim
And I think ultimately the utility of young people is to question deeply held beliefs, is to be heretics.
Amjad Masad4:07
There's a lot of ideas that don't come out of Silicon Valley because there's a monoculture here.
Amjad Masad5:11
And I think a lot of starting startups is premature optimization.
Amjad Masad17:35
I tried not to start rapidly. Like actively not to start it, but it kind of exploded into a startup on its own.
Amjad Masad18:37
In Silicon Valley, I think it's. Probably 75% to 100%, depending on who you're talking to and who you are.
Gagan Viani29:53
We don't have grades. We don't have tests. The sort of amount of weight to push against should be a lot less.
Gagan Viani32:01
did you actually follow a thread all the way through? That's always really impressive
Amjad Masad44:20
Figures
| Probability a verbally committed nonprofit donation actually arrives | 20% to 50% | 28:51 |
| Probability a verbally committed Silicon Valley donation actually arrives | 75% to 100% | 29:53 |
| Investment a VC gives Stanford students (as recounted by Amjad) | $1 million | 18:37 |
| Length of the exploration period Gagan advocates | Two years | 15:33 |
Glossary
- premature optimization
- In programming, optimizing code before you know what matters; Amjad uses it as a metaphor for starting a company too early.
- Holacracy
- An organizational form that abolishes management hierarchy in favor of autonomous roles; Medium and GitHub tried it and failed.
- self-driving company
- Amjad's term: agents take on bureaucratic processes and the company ultimately runs itself.
- vibe research
- Amjad's name for AI doing machine learning research itself, by analogy with vibe coding.
- virtue ethics
- The Aristotelian ethical view: act in a way you would not be ashamed of whether or not anyone is watching.
How to listen
Young people agonizing over whether to go to college or start a company now, plus founders, investors and anyone hiring, funding early-stage projects, or trying to understand how AI changes organizational forms.
The sci-fi novel recap at the start, 0:00-2:04, can be skipped; go straight to the argument starting at 4:07.