I am starting to think this world rewards people who are crazy and punishes people who do what makes sense.
The sensible move right now is to flinch, and the flinch is everywhere once you start looking for it. Layoff memos. Token leaderboards going flat. A lab missing a quarter. Some guy in the group chat announcing that the bubble popped, the institutions are cooked, democracy is breaking, education is broken, pick your Gutenberg moment, we are in the trough, hide under the desk. Get small, protect the baseline, wait for spare time, wait for Australia to catch up, wait for the hallelujah moment to happen to somebody else. That is delusional pessimism, it is the majority sport of our generation, and I think it is the wrong read of the year we are actually in.
I promise this will not be motivational slop.
The new attitude
The new attitude is not "believe in yourself when no one else does and everything works out." That sentence is annoying because it skips the part where luck is required and the part where most people never put themselves in a position to get lucky. The new attitude is simpler than that. Be delusional about the destination. Be paranoid about the pipes.
Delusional means you shut the door on the thought that this is already over. You do not let "what if it collapses" sit in the room all day like it pays rent. You move like the next decade is the opening, not the credits. You couple with AI and ride along with it, and you do not spend your one life as a stochastic parrot for other people's fear.
Paranoid about the pipes means you do not LARP as God while the corpus walks out the door. Building secure AI systems that do not lose people's data to criminals is a good mission, and there are years of work in it: inventory, risk, controls, evidence. AI governance is not a vibe. It is fifty years of "the demo is the product" getting unlearned, and a new model getting learned in its place, where the only proof that counts is a system that runs under load, with logs and an owner.
Most people hear those two sentences and think they cancel. They don't. The destination is abundance. The pipes are how you don't get taken apart on the way there.
If you only do the delusion, you become an Abundance Summit extra. If you only do the pipes, you become the chud who thinks about failure all day, takes the safe route, and watches other people win while telling himself that for him it is different. I have done that version of myself. When nobody was calling me crazy I had the job that made sense, the credentials, the path. I also had the tax you pay with your soul when the work is real and the upside belongs to somebody else.
The route that makes sense no longer makes sense. That is the new crazy.
Why he is right
There is a guy on YouTube named Poly Walnuts who made a video called "call me crazy." He decided to be a nut job on purpose, wrote it in a journal in February 2025 so that he could go back later and see the exact date, and he says the scary part of moving crazy is that you don't have evidence yet. Anyone can get lucky once. The work, the consistency, the pit in your stomach, nobody sees that part. They see the dashboard a year later and call you lucky. So you put yourself in a position to get lucky.
There is also a room full of moonshot people who will tell you the same thing in different clothes. Couple with AI or be left behind or both. Don't fight missionary battles. Leapfrog. Leave slack in the calendar on purpose. Super intelligence, if it shows up, invents whatever comes after it. I am not here to sell their products. I am saying the attitude is correct.
He is right that the generation's problem is not an ego that is too big. It is an ego that is too small. You have an idea and you shut it down before you try it, you take the rational route with school and work and the thing that keeps popping into your head, and you wait for the calendar to clear, and it never does.
He is also right that other people cannot see your vision. Your friends, your family, the CEOs who are too busy to think, the Australian operator who feels behind in an exponential age. They are not villains. They are looking forward, trying to connect dots that only connect backwards, which is the thing Steve Jobs already said and is still some of the realest material in circulation. You trust that the dots will somehow connect. At the time it feels like the end of the world. A couple of years later it looks like it was written by an author.
And he is right that luck is required, and that needing luck is not an argument against going. You cannot put yourself in a position to get lucky if you don't have the delusional self-belief first, so that if it happens, you're ready.
The moonshot people are right that this is not a prediction-machine story anymore. Reasoning models got reinforcement-fine-tuned on hard problems until "stochastic parrot" stopped being a fair description of what they do. They produce novelty now. Inventive was already in the latent space, and there is enough compute under the hood that the self-supervised task has a disgusting amount of runway left. Transformers are already a ship of Theseus. The compute substrate will move, photonics and then whatever physics the current stack invents and explains to us like we are the slow kids.
They are right that institutions are going to have a painful stretch. Healthcare, legal, IP and education are still running like the 19th century, twenty Gutenberg moments are arriving at once, and humans hate change. That part is unpleasant. It is also the only mechanism that clears the ground, and abundance does not arrive while the old systems are still standing.
They are right that overregulation is a way to lose a race nobody gets to pause. A hundred and fifty thousand people die every day. Super intelligence as an antidote for medicine is not a TED Talk if you have watched a blunt-force pharmaceutical sit on a diagnosis for thirty years. Hands off does not mean "no pipes." It means do not hamstring the thing that might work because you are performing safety.
I used to think I did not want a crazy career. That was the chud move. This applies if you are in Defence, in cloud, sitting in a room in Da Nang drinking sponsored coffee, or in a mid-career body with a mortgage-shaped set of constraints. When you are young you can take more risk. When you are older you have more leverage, so use it in a heavy way. What you know is not common. Domain expertise is not a consolation prize. It is the proprietary operational data of a life.
The collapse means nothing
People keep asking for the collapse like it is a destination.
The model stock dumps. A lab has a bad month. An agent does something stupid. A grid flickers. A bank gets poked. The collapse is weather, not climate, and the P(doom) crowd gets a week of content out of every squall. Eliezer's cousins tell you to be anxious, the "upper" podcasts tell you to be visionary, and you are supposed to pick a team.
Every technology has caused harm on the way in. Books in the 1500s were going to destroy memory, create the illusion of knowledge, overwhelm people with information, spread misinformation and corrupt the young. Same pattern. FDR had to say the only thing we have to fear is fear itself because fear was the product. My grandfather's version was bomb shelters. The immune system of a culture does this every time.
Defensive co-scaling is already the industry. Zero-days exist because humans wrote the code, and the rewrite is coming. There will be rough spots, two to eight years of stupid if we are unlucky. Civil unrest is not a fantasy. Young men with no job, no family and everything too expensive is a pattern older than GPUs. None of that is the same sentence as "it is over."
The collapse of a seat is not the collapse of the era. A salary is a per-seat license, and when execution gets cheap the company buys fewer seats. That is arithmetic, not apocalypse. If you owned the seat it feels like the world ended. If you own an outcome, a spec, a relationship, a system that still runs on Monday, it is the biggest force multiplier anybody in our line of work has been handed.
AI winter, if it shows up, is a funding weather report. The self-supervised task does not care. The baud rate of a human brain does not care. Your own AI will eventually filter and summarize better than you can, and it does not care either. BCI, full-dive, uploading, pick your sci-fi. Even if half of that is late, the interface shift has already started, and you either couple with it or you get left behind. Both can happen in the same world. Both probably will.
So the collapse means nothing in the only sense that matters: it is not a reason to go back to the sensible route. You were already in a pretty dark equilibrium if that was the plan. A job you hate, not enough money to justify the tax on the soul, CEOs too busy to think, emerging markets waiting for permission from Harvard MBA ghosts. What's the worst that can happen. You have already seen the worst case, and it was being sold to you as normal.
If I lose the particular shape of this decade, if a channel gets copyright-struck by a bot, if a product dies, if a room empties, the honest line is still this. At least I got to try. At least I had a voice for a minute. Nobody was calling me crazy when I had nothing to show for it, and I hated that.
It is the year 2000
This is the part people cannot feel because they are inside it.
When the internet showed up, people did not adopt it for fifteen years, and then it was the interface to everything. AI is going to be the same way. You cannot force the horse to drink, and you cannot build your life around the people who will never cross the line. The world is just going to pass them by.
Year 2000 is not "we made it." Year 2000 is the interface being obvious to a few people and optional to everyone else, with the optional people thinking they still have time. Search was dumb, commerce was embarrassing, and the infrastructure was kind of mostly there. Leapfrog was available to anyone not trapped unlearning a prior era. The people who waited for spare time never caught up.
That is this.
LLMs are not the end of the S-curve. They are a language modeling task with infinite longevity, next token, next bit, missing bit, solve the mysteries of the universe, or at least stop pretending the current notebook is a brain. Hardware is getting fundable again because the products have atoms. Energy per unit of intelligence will fall while the absolute amount of energy rises, and somebody will call that a bottleneck until it isn't. Longevity, education, disability, caregiving, the things that lag in the physical world, they start with planning and agents and then they roboticize. None of that is a finished story. It is a beginning that already has enough surface for a life of work.
I wish I was goddamn 18 today. I also am not. I am the other end of that spectrum, the one who found the energy late and still feels it. The playbook is not 80-hour weeks in a basement. The playbook is leverage. Describe your actual life to the model, the obligations, the kids, the constraints, the domain, and brainstorm on the margin: how much time, how much money, what can I do. Build your own Skippy at night. Fabricate a little cash flow so the market knows you exist. Find five people you want to be the average of, geographically or not. Cafe Cursor was never about the perfect harness. People come to those rooms looking for a software factory. The factory is an attitude plus guardrails. The harness does not matter.
Australia can feel behind, and feeling behind in an exponential age is terrifying. The good news is that you can leapfrog really quickly, because you don't have fifty years of the old model to unlearn if you never fully believed it. Distance is a playground. Tinker. Leave ten or twenty percent of the calendar with nothing on it, or you will overschedule yourself into 2019 forever.
Don't fight missionary battles. Be there on the day they get clubbed over the head by the hallelujah moment. Build the thing that is ready when they walk out of the room a different person.
Pipes
Max delusion is the obvious answer. I mean that.
Not as a pitch, but as a decision you make once, on purpose, the way Poly Walnuts wrote it in a journal so he could go back and see the date. I am going to be delusional. I am going to ignore the part of the culture that is paid to narrate decline. I am going to treat this as the start.
The correction that keeps this from turning into a 5150, the California code for an involuntary hold, is the pipes.
God does not fail is a useful LARP for a presentation you didn't sleep for. It is a stupid operating model when the corpus belongs to someone else. Your subconscious does not know the difference between "I can do this" and "I already did this." Criminals do. So do models, when you leave the corpus on the public internet and call it innovation. A true genius is supposed to be humble and recognize that the gifts were given. Humility here is not shrinking. Humility is: I can do it all, and I am not going to become the guy who cannot be trusted with the keys.
So the mission that does not need a landing page: secure AI systems. Do not lose people's data to criminals. Governance with years of runway. That is not the non-ambition route. That is the work that has to exist if the abundance era is going to be livable. Defensive co-scaling on purpose. Red team like you mean it. Hands off the reflex to regulate the future out of existence, hands on the actual breach.
Satisfaction will not come from a number. You hit the goal and you move the goalpost. You get on a treadmill. Take the step back anyway. Gratitude is still the key to everything, and ambition without it is just a louder chud. I would rather complain about winning than complain about losing. I have not made it. I am far from far from making it. The mindset compounds, though. A little evidence that you can win, and suddenly more people treat you like you can. Nothing about you changed except the story they are allowed to believe.
Connect the dots looking backwards. Thank God for the ugly years. Thank God for the boss, the contract, the city, the accident that forced the quit. At the time it is bad luck. Later it is the only reason you had a chance to be crazy.
Manifesto
- The world rewards people who are crazy and punishes people who do what makes sense. Act accordingly.
- Our generation's problem is delusional pessimism, not too much ego. Ask yourself whether your ego is big enough.
- Max delusion is the obvious answer. Shut the door on collapse-as-identity. Do not let loser thoughts pay rent in your head.
- The AI collapse means nothing. Seats collapse. Eras do not. Weather is not climate, and harm is not the end of the pattern.
- It is the year 2000. Optional to most people, the interface to everything for the ones who couple. You are not late. You are early and acting like it is closing time.
- Couple with AI or be left behind or both. Both will happen in the same world. Pick the first one on purpose.
- Don't fight missionary battles. Be there for the hallelujah moment. Leave slack in the calendar on purpose.
- Be delusional about the destination. Be paranoid about the pipes. Secure systems, no corpus for criminals, years of governance work. Demos expire.
- Luck is required. Put yourself in a position to get lucky. What you know is not common, and older is leverage.
- Other people cannot see your vision. You cannot fully see it either. Trust the dots. Connect them later.
- Optimism should be high. If this works, the future is actually good: intelligence, energy, longevity, all of it cheaper. Rough spots on the way. Still go.
- When nobody is calling you crazy, check whether you have anything to show for your life. If the answer is a job that makes sense, that is the new crazy. Leave.
I seriously think you have to be crazy. I think it is crazier not to be crazy in 2026.
Peace.