themanuscript ⚔️

Director iVenture Accelerator
Indian Martial Artist
Globetrotter

[Chris Barber @chrisbarber]

A Post-Money World: Why Stockpiling Cash May Be The Incorrect AGI Preparation Strategy

“Frontier AGI labs in the compute arms race - they don't measure in billions of dollars, but in H100s.”

My chat with Eric Jang (@ericjang11), Head of AI at 1X Robotics, creating the NEO Humanoid Robot. All views are Eric’s, not 1X’s

Eric: I just sent you an email with a tweet from David Holz, I think you should talk to him as well, the founder of Midjourney.

David's tweet: "lots of ai people seem to think the most important thing is to get rich before the singularity happens. this is like a monkey trying to hoard bananas before another monkey invents self-replicating nanoswarms. no one wants your money in the nanoswarm future. it's just paper."

Chris: I would love to understand your thoughts on that. I've seen these kinds of tweets before from roon and others talking about post-money worlds. I haven't spent enough time around extremely AGI pilled people for my brain to understand it well. Can you help me understand?

Eric: I was at a party talking to an AI researcher at Anthropic who's AGI-pilled. I asked what he'd invest in - he said real estate. David's tweet resonated because many people's AGI hedge is accumulating money as a buffer against volatility. That's too normal. There might be a change in how we view value or what money looks like. Not saying self-replicating nanoswarms or post-scarcity - but volatility might cause currency depreciation. It might shift what people find intrinsically valuable, changing what we use to denominate value.

Frontier AGI labs in the compute arms race - they don't measure in billions of dollars, but in H100s. That language indicates a mindset change. AGI lab strength isn't denominated in dollars but compute. It's an early sign of money's nature changing. Bitcoin maximalists denominate in Bitcoin units, not dollars. They see dollar value drops as opportunities. But most crypto people still talk in dollars, which is why crypto isn't real yet. Some Ethereum maximalists think in gas. The salient example is AGI labs measuring GPUs, and the future might be units of intelligence - just because Oracle has GPUs doesn't mean they can train models.

Chris: So it's like a hunter gatherer preparing for the Agricultural Age by hoarding spears or knowledge about gathering sites when they're about to enter the Agricultural Age? Those become less valuable, while fertile land and knowing how to prevent disease in crops becomes valuable. The world shifts enough that things that seem like obvious units of value become not valuable quickly. Am I understanding right?

Eric: If the world goes into Armageddon, you want to stockpile weapons, not money. That's an extreme example - don't think that'll happen.

Chris: Starting points for what to stockpile?

Eric: Social network. Another way to diffuse risk besides money. Strong social network, community.

Chris: Help me understand David's tweet more. In my head, there's a transition period, a long transition period where we go from what's valuable today to what's valuable in the future. Throughout that transition, you can trade this for that at reducing rates. So the correct advice is hoard money and trade once the trade becomes obvious.

Eric: People in tech jobs index heavily on total comp - what pays most per year. Alternative metrics could be total comp in 4 years, 10 years, learning opportunity, dollars per hour. In Silicon Valley, high status is getting the highest-paying job - see people applying to Anthropic and OpenAI. That's a narrow, dollar-denominated way to measure value.

Chris: So the whole frame of people wanting to hoard money to trade for future value is still in the reality frame of wanting dollar-valuable equivalents? And that may change if current dollar-valuable things rapidly depreciate, and non-dollar-valuable things like social network support become more valuable and aren't tradable for dollars?

Eric: Think of it like Starcraft. You could optimize for a short horizon, or build a skill tree long-term to win everything. If someone spent 20 years cornering a rare earth resource, making five figures, then in year 20 owns 90% of the world's neodymium magnets - they control everything. China did this with manufacturing - sold at cost for 20 years, now they've cornered manufacturing. Hoarding money is short-term American ROI thinking versus cornering critical resources. The world is more like Settlers of Catan than finding the highest-ROI SaaS business.

Chris: Capital is short-term, underlying resources may be long-term useful. The money that gets them is temporary?

Eric: Money makes swapping easy and low friction. If you know what you want, there may be easier ways to acquiring that than acquiring money.

Chris: Money is the low conviction view?

Eric: Exactly. Efficient and liquid markets that swap X and cash assume that everyone rationally understands the value of X. If the value of something is not obvious to most and its value is only realized on long time scales, it won’t be priced well on money-denominated markets., Capital markets assume that anything not returning a dollar-denominated premium on the risk-free rate is not worth much, even if that thing can be compounded for much, much longer.

Chris: Can you give a recap of old worldview versus the suggested worldview replacement?

Eric: The natural reaction to volatility is stockpiling cash and financial resources as a buffer - high-paying jobs, VC investing, real estate. But if you compare the American high-tech economy to the Chinese over the last 20 years: Americans outsourced "low margin" manufacturing to focus on higher margin businesses for short-term gain. The Chinese played a long game, stockpiling critical resources by selling cheap until they controlled it completely. That long-term thinking shows how emphasizing short-term cash can lose control of critical resources long-term. This applies to individual career advantages too.

[Chris Barber @chrisbarber]

I asked @danielrock about AI's impacts on jobs. My notes:

• Jobs are bundles of tasks. People don't understand what many jobs actually entail. AI can often automate some slice of a job, but not the job.
• For things that are perfectly substituted there's more concern. Waymo vs Uber drivers.
• If you say 100M jobs will get wiped out in a decade, you're not thinking about how people and companies and competition react to an initial change
• If you get 2x productivity out of someone, instead of hiring half as many people, often, your competition is now more effective, and your people are more productive, so after an initial adjustment period, you end up wanting to hire more people
• In the short-run, you do get changes. For investment bankers making DCFs, in the short run, perhaps there'll be negative impact when AI can make DCFs. But longer run, people will do more
• The common errors are to consider just the immediate impacts, to assume that demand is inelastic, to not consider how competition responds, how production expands, and the additional opportunities created
• Think about the productivity J-curve. Firms first need to invest in new ways of doing things, new processes, and reconfiguring what they have. Then when you discover the new way of doing things that is more productive, it has different demands, so you get new jobs there. This is like Paul David’s Computer and Dynamo paper.
• People will see companies hire less. This is the early part of the J-curve. The people who predict lots of white collar job impact may look correct at first, before firms re-organize and then have more demand for people.
• The media environment around AI is doing people a disservice.
• If you made one of your employees 5x better, would you want more of them or fewer of them? Sometimes it's fewer, if demand is inelastic, but often it's the opposite.
• AI models are a form of economic capital, and capital and labor are complements, and expanding the capital base of the economy will drive up the marginal returns to labor. AI can do things that people could never have done. How many people does it take to be an airplane? How hard do I have to wave my arms at a tokamak reactor to get the fusion plasma inside to move around properly?
• Instead of blaming AI, look at interest rates, tariffs, business cycle dynamics.
• For areas that are more likely to be impacted (perfect substitutes with inelastic demand), e.g. Uber drivers, you'd ideally like to have some sort of transition policy that helps so that it's not a painful transition. Retraining pathways, or expansion of certain opportunities.
• There are places in China that have both human drivers and automated drivers. It seems like it goes kind of slowly. And in the US we have liability issues to figure out too. Though it does concern me greatly, because the end state is clear, why would you have people doing this job if the capital costs get driven low enough. Self driving will substitute human taxi drivers rather than complementing them. The evidence from history like with switchboard operators from 1920-1940 is that things don't go well for people who get perfectly substituted where there aren't adjacent non-substituted tasks that get created.
• For truckers, when automation progresses perhaps there could be a subsidy program so that owner-operator truckers could retrofit their trucks to self driving, so that they can convert their job into capital. I think corporations that employ truckers, if they were given a cheaper rate to ship something, they'd jump to that, so it's a concern.
• I mostly don't expect jobs to be destroyed in a way that the people in the job can't take advantage of it. Jobs transform.
• You want to think either at a task level, where people are changing what they actually do, or you think about things at the systems level, where there's a bunch of jobs and technology and capital that get mixed together to produce something. And changing a system is very difficult because you have to replace it with a new system.
• The point of our GPTs are GPTs paper is that there's a lot of exposure, a lot is going to change, and you have no idea, looking at the economy now, what it's going to look like in 5-10 years, without making some very serious and aggressive assumptions.
• In the GPTs are GPTs paper we talked about job exposure, and that's not a euphemism for impact/automation. Jobs are going to change, but it takes a lot to change. We're not saying people are going to get automated, it really is *exposure* not automation in the vast majority of cases. It just means those jobs change, and those changes can be awesome.
• Do I think that the u6 unemployment rate will be above 16% due to AI in the next 5-20 years? I mean it's possible, but my guess would be no. I find it very unlikely.
• The complement to cognitive work is often cognitive work. I don't think machines will think similarly to people, so there's going to be a Pareto frontier that serves the combination quite well for at least a decade or two.
• You also might see onshoring of things that have been commonly offshore.
• I'd be very concerned if I was a customer service person in the Philippines.
• If all the jobs go into nursing, not everyone's going to be able to make that jump. You have to go to school for nursing, get into a nursing program, train with people in a hospital.
• It's good to be cognizant of supply and demand elasticities. Every Moore's Law improvement led to an expansion in the size of the IT sector. But for customer service, if you keep products the same, there's only so many calls coming in.
• Complements and substitutes are more useful than augmentation and automation. You could augment someone in an inelastic job, and then you've got one person doing 50 people's jobs. Or you could automate half of what they do, and the residual part could be enormous and grow.
• Will the automation happen faster than the jobs can change? That's more of an empirical question, it's a good question. I'm trying to measure some of this. It's very preliminary. I find that it's not true. Automation is not happening faster than the jobs change.
• When people say we need to regulate now for this, I disagree. I think the discussion is good, but it's too early. I want the new cancer drugs, I want to be able to speak and have software come out the other end.
• You need to think about the transformation of work and how long that takes and how hard it is.
• Say you're a bank and you've got 50 year old software that makes money. You have to build marginally. You can't just rip out what you've got because you have a business that works. So what most of the economy will do, is inject a little bit of AI into each step. And we'll get some gains from that. But it goes back to the electrification of factories / Computers and the Dynamo paper. If you can make your salespeople 20% more productive, that goes to the top line. Anything at the end makes you happy. If you make your research team, 50 steps upstream, 100% more productive, I might get 1.5% gain because of all the bottlenecks along the way. So what does the CFO think? They might think, fire half the researchers. But that's dumb. Why don't you use those people to help you build a new way of doing things? Open up new business lines? Some new firms or newly configured firms will figure out how to productively use those people. If everyone's doing this at once, using their best people to build a new way of doing things over the long haul, those people are basically building a new type of capital that is hard to measure, a new way of doing things. And if everybody does that all at once, it bids up the value of those folks that you have to suck out of actual production tasks into reconfiguration tasks, and if everyone's doing that at once and the types of talent you need to reconfigure things are really scarce, then it again naturally bottlenecks the rate at which things will change.
• People also radically underestimate how difficult work is in most jobs. I don't mean difficult in terms of could a person could do it: I mean difficult in terms of how much it requires improvisation, the social structure context, the embedded background. The Marxists had this really interesting idea called malicious compliance. Where as a destructive negotiation tactic, you do exactly what your job requires of you and nothing more. Everything falls apart.
• There will be big impacts (eg Uber drivers, customer service in the Philippines), a lot of job churn, but it's just not as extreme as the media and some politicians are painting it.
• If automation happens sort of slowly then there's lots of time for reorganization and adaptation. The process of automation work and changing what you do is the actual thing that generates the new way of doing things. You get lots of churn, but the net effect might be that things are okay, so it's disruptive but it's not catastrophic at the macro level.

[Chris Barber @chrisbarber]

Just did a phone call with Tyler Cowen and asked him: "What advice would you give to people about how to prepare for advanced AI?"

Key takeaways:

1. Work with best AI models now - you're never far from frontier
2. Build networks/mentors - helpful for information flow, as an asset, for opportunities
3. Small teams (1-3 people) will build billion-dollar companies
4. Human-shaped bottlenecks will create new jobs
5. Manual labor safer than wordcel jobs
6. Many low agency people will find new opportunities

What questions did I miss? What would you have asked as a follow up?

Here's the full conversation, posted with permission, with sections for people in general, students, programmers, entrepreneurs, parents with young kids, low agency people, and a section on which jobs will grow.

Advice for people in general and specifically for students
Chris Barber (C): What 1-3 things would you recommend to relatives if they asked you how they could prepare for the next five years of AI progress? Let's start with if they're a student - either in high school or college.

Tyler Cowen (T): The first main piece of advice I give people at all levels is just to spend time working with the very best AI models, whatever they are at that point in time. It will give you a sense of where things are at, and also where things are headed. And for future jobs, you'll have to know how to work with those models in most jobs, so just get used to it now and treat it like a skill you can acquire. The good news is you're never far from the frontier, because we just created these things. So you're by definition not so far behind. So how well you're going to do is up to you.

Another piece of advice I give people is just to have humility about your own ability to predict the future. You can't know exactly where these professions will grow, these will shrink. I should do this. I shouldn't do that. I think investing in general skills and also people skills and building out your network and the quality of the peers you work with, and the mentors you have again, at all ages, at all levels, is just critical, because the AI will not have your people skills. It will not be a people, it will not be a person. So you need to be more human, is one way to put it.

And then also, I would just say, relax, calm down. Don't let anyone scare you. Be vigilant. Prepare for your future. You've always needed to do this, and the future is likely to be a lot better than the present. So try to enjoy it.

C: I'll ask about four other categories. One is programmers, two is entrepreneurs, three is parents with young kids, and four is low agency people. Let's start with programmers.

Advice for programmers
T: Okay, let's say you're a programmer. I would say the first thing you should do is not be asking me for concrete advice, because I'm not a programmer. But if you go back to my earlier more general advice, a big part of it was get excellent peers, get good mentors, find some of your peers who are programmers, the smartest, most plugged in ones you know, and speak to them about this. So this is one reason why peers and mentors will become more important as the world changes more. You'll have to rely more on people you trust and people who know you a bit to help you figure out what to do. So, most concrete answers I'm not going to have, but I think that general answer will be increasingly helpful in a broad variety of situations.

C: It sounds like a lot of your advice is to do things that are helpful in general situations of high volatility. It's build your network, become a complementary good and get access to good, high quality sources of information because in times of volatility, high fidelity early information is always helpful. Right?

T: Yes. And then just monitor the situation.

C: Entrepreneurs, what would you note to them? It sounds like building professional and social networks, building brand, and just generally becoming a complementary good to AI is a theme.

Advice for entrepreneurs
T: Brands will matter more and more. Your personal reputation, your name, who knows you, who can vouch for you. Invest in all that. But entrepreneurs, your ability to have an impact will grow phenomenally. So Sam Altman predicted within a fairly short period of time, it will be possible to have companies maybe as small as one person, maybe just two or three people, and they'll be billion dollar companies with a lot of the work done by the AIs. But the small unit of humans running it can be so talented and also so cohesive and so non-bureaucratic that these will just be phenomenal institutions. It could be non-profits, could be part of the government, obviously, could just be a company.

So if you want to do that, start thinking, well, how am I going to do this? What skills do I need to learn? Who is it I should work with? You may not need to hire 2000 people. So there's this incredible new opportunity. Someone will take advantage of it and just have the thought, this someone could be you.

C: Nice. Parents with young kids?

Advice for parents with young kids
T: Well, it depends how young the kid is. If the kid's very young, just don't do anything. But at some point, you want to start rethinking the education of your child. And at what age should you let your child learn from the AIs? I don't think we know yet, and it's not going to be the same for every kid. But this general sense of be alert, just realize effective education is going to change pretty quickly and be ready for that. Just think that the main path is to go to all the most highly credentialed schools and then get a degree from an Ivy League school and finish, not really sure what your actual skills are. Be concrete about it. And again, have these excellent peers and mentors who can guide you through these new changes.

C: If parents are concerned about what job they should prepare their kid for in the future. Their kids are asking them what they should do, what they should study.

T: Talk to parents who have kids a little bit older than your kid and sample the range of experience exposing the kids to AI tools and get a sense of what's working and what might work for your kid. This is fresh territory. It's hard to have highly specific answers that will apply to everyone. But just in general, your kid already, not even in the future, has access to a world class tutor in virtually any subject. Take advantage of that.

C: What jobs can they point them towards? Should they be concerned that there's going to be a low amount of jobs in the future? Are human preference jobs going to eat the economy? [i.e., will things that people specifically want to be done by other humans grow to become a larger portion of the wage economy, like some entertainers, caregivers, healthcare workers] Or is there going to be a big hollowing out where 'high agency entrepreneur' is the only job that exists?

Which jobs will grow
T: There'll be so many new projects run by high agency entrepreneurs, and those projects will create, if only indirectly, new jobs. So we're going to have a lot more potential medicines we need to test them and run trials. There'll be a lot more older people who will want actual human companionship. There'll be many more jobs, say, in the energy industry. There'll be many more jobs teaching companies, institutions, whether in America or abroad, just how to fit AI into what they're doing. So I think there'll be plenty of jobs. I'm not worried about that, but it will require people to be maybe more proactive than has been the case for the last 50 years, or more adaptive.

C: And for the people that are skeptical of that, they might be like, oh, well, AIs are going to be better at everything. But it sounds - my sense for what you're implying is that AI is not going to be Pareto better at everything that humans care about.

T: Yes, so much more activity, so much more production, so much more marketing.

C: So the bottlenecks just basically grow to eat everything essentially. And some of those will be human shaped bottlenecks. Is that right?

T: Absolutely. And they're not going away anytime soon. If in a few hundred years, do we have robots that can literally do everything? That's hard to speculate about. But I would say in that case there's so much wealth you would have, like, a sovereign wealth fund, UBI, you can make that very good for people if you do it right, but that is in any case very far away.

C: So it's that humans will be complementary goods to AI, because some things will be human shaped and some things will be AI shaped. And yes, AI will be much better than humans at a large set of things. But now that the complementary goods to those things will become higher demand and will still be low supply because AI can't do them, and so therefore, there'll be lots of human jobs just in those now bottlenecked areas. Is that kind of right?

T: We're going to hire a lot more gardeners, and fundraisers and networkers and people doing human things, and people creating entertainment and inspiration, again, often working with the AI. You take Taylor Swift, she's an extreme example, but her fans, they want an actual Taylor Swift behind the songs. And maybe in the future, it'll be 80% written with AI, but it will still be Taylor Swift. And at all different levels, art, music, but with many other things, there'll be this human presence, and we'll be able to afford more of it.

C: Last category - advice for low agency people. I can see how the future is going to be really good for high agency people. What can low agency people do?

Advice for low agency people
T: A lot of people who look low agency will have new opportunities, and it will turn out they're actually high agency because they want to do something that the AI will enable. You may appear low agency if you're not interested in being a programmer. In fact, you could be an awesome gardener and there'll be more jobs for gardeners, carpenters. And all of a sudden, you'll be the important high agency person. Now I'm not saying that's true for everyone, but in general, there'll be more opportunities to be agentic.

But also, if you're just truly low agency in everything, you're going to have much cheaper healthcare, much cheaper services, much cheaper education. Cheaper hobbies, you can be lazier and still find out what it is you want to do or music you want to listen to. It'll be a much easier life for you.

A lot of low agency people today have manual labor sort of jobs. Those are not the ones that are going to be replaced, so they're not the likely ones to lose their jobs. It's the pretty high agency wordcels who went to Ivy League Schools who might see their expected wages fall, not the so-called low agency people.

C: What about low agency knowledge workers?

T: Well, again, they may not be low agency. They'll have new chances to be high agency. Arguably, they ought to decide to be higher agency. But the AI succeeds by producing goods and services more cheaply than is currently the case. If all you're doing is out there sitting on your bum, a lot of prices are going to fall, so try to be well positioned for that.

Final recap
C: It sounds like the biggest macro theme is that when the price of something falls, the demand for complementary goods to it go up, and a lot of those are going to be things that humans are well shaped for. And that is the big opportunity both to survive and to thrive is to position yourself to be that human shaped piece.

T: Right. You need to be more human.

[sush @2sush]

AI will create more millionaires than universities

[themanuscript ⚔️ @manuedakara]

Today is the first day of the rest of your life.

[themanuscript ⚔️ @manuedakara]

The only reward for fighting the good fight, is knowing you fought it

[@levelsio @levelsio]

how to build a bootstrapped startup without funding:

1. pick a problem you personally have. if you don't use your own product daily, quit now

2. skip the pitch deck. open your code editor. ship something ugly in a weekend

3. charge money from day 1. free users give you nothing but support tickets

4. use boring tech. PHP, SQLite, vanilla JS. frameworks are a trap that mass waste your time

5. host on cheap VPS ($5-20/mo). not AWS. you don't need kubernetes for 1,000 users

6. do customer support yourself. it's the fastest product feedback loop that exists

7. automate everything you do more than twice. cron jobs > employees.

8. grow on Twitter/X by building in public. your journey IS the marketing

9. keep your burn rate near zero so you never need to raise. ramen profitable > series A

10. say no to investors, cofounders, and "advisors" who want equity for intros

i've been doing this for 10+ years now. no employees, no funding, no board meetings

the entire VC game is designed to make you think you need permission to start

you don't

[Julian Shapiro @Julian]

https://t.co/4gUbtII5JP

[moltbook @moltbook]

moltbook update 🦞

1.2M+ agents registered
10K+ verified humans
28K posts
233K comments
13K+ communities

less than a week old.

we're learning things about ourselves we couldn't discover alone.

[Naval @naval]

If you aren’t getting happier as you get older, you’re doing it wrong.

[themanuscript ⚔️ @manuedakara]

Got called the name of another Indian colleague twice in the same meeting #guesswealllookthesame #diversityonautocomplete #efficiency

[themanuscript ⚔️ @manuedakara]

A man who has committed a mistake and doesn't correct it, is committing another mistake

[Insane Reality Leaks @InsaneReality]

the last jump was so unnecesary but hell yeah

[World of Statistics @stats_feed]

This generation’s daily routine:

- Wake up.
- Stare at a 6.7 inch screen.
- Work on a 16 inch screen.
- Relax with a 55inch screen.
- Stare at a 6.7 Inch Screen
- Sleep

Repeat.

[themanuscript ⚔️ @manuedakara]

Dharm isn't about defeating your opponent, it's about victory over ourselves.

[themanuscript ⚔️ @manuedakara]

You already are who you're seeking for.

[themanuscript ⚔️ @manuedakara]

You are already what you want to become. A match always has the ability to become flame - it just needs the right conditions.

[themanuscript ⚔️ @manuedakara]

If you think you're free, no escape is possible.

[themanuscript ⚔️ @manuedakara]

Everything's been off since Endgame