second pour.

no rush.

sit. stay a while.

open your feed and the story is that ai got too expensive. uber gave five thousand engineers claude code in december, burned the whole annual budget by april, then capped everyone at fifteen hundred a month. ninety-five percent of corporate ai pilots show no measurable return. i felt a smaller version of it myself last year, watching the perplexity and claude and cursor charges stack up, thinking is this even worth it.

here’s the strange part. at the same time, the price of intelligence is in free-fall. a model you’ve never heard of serves the same answer at a tenth of the famous one. so which is it, more expensive or radically cheaper?

both. and uber’s ceo said why in the same breath: “we use the more expensive models to explore. once we scale, we bring in cheaper or open-source ones.” that one sentence is the whole lesson.

the lesson this week.

the price of intelligence keeps falling. that part is certain. what’s not: cheaper has never meant less spending, it means more. so the market splits in two. the frontier keeps the work where intelligence still matters; cheap, good-enough models take everything else.

the syllabus.

six questions, in the order they come up. this isn’t a sitting. do one every couple of days and let it settle before the next, two weeks if you do it right.

i. why does ai suddenly feel too expensive?

the bill is the problem, not the price.

start where the conversation is. uber gave 5,000 engineers claude code, usage doubled in two months, the budget was gone by april. on all-in, jason and chamath name the scary version: at three hundred a day per agent, the token bill starts chasing the salary. it feels like the bill is the problem.

the material.

• all-in — when ai tokens cost more than your employees ↗ listen

• aakash gupta — the uber adoption curve ↗ read

things to notice:

• the panic is about the bill (total spend), not the price (cost per answer). hold those apart.

• uber didn’t quit. they capped it and kept going. the spend exploded because it worked.

sit with these. then take them to claude.

• what do you pay for ai each month, and have you ever added it up the way uber just had to?

• if the most disciplined operator in tech can’t scope its own ai bill, whose revenue model is underwater?

• when the token bill passes the salary it replaced, what happens to that job?

ii. wait, isn’t it getting cheaper?

it is, faster than almost anything ever has.

now the other side. epoch tracks the two clocks: about 3x efficiency a year, and the cost to run a model at a fixed quality halving every two months. tunguz puts a face on it, deepseek cut training cost ninety percent and hit top performance at a fortieth of the price. his read on demand: it won’t shrink, “it’ll explode.”

the material.

• epoch ai — trends ↗ read

• tomasz tunguz — the ai cost curve just collapsed again ↗ read

things to notice:

• the honest unit is price for a fixed level of performance. last year’s best is nearly free this year.

• the cheapness is cleverness, not selling below cost. it’s real. (the frontier does the opposite, see v.)

sit with these. then take them to claude.

• which ai task cost you real money last year and is basically free now, and did you re-price it?

• if a fixed level of performance keeps getting cheaper this fast, whose pricing power is melting?

• what becomes worth doing the moment it costs a hundredth of today?

iii. what are you actually paying the frontier for?

a four-month head start. that’s the whole premium.

this is where the lesson lives. epoch measures the gap between the best open and best closed model: about four months. yash patil, who built models at openai, says moving your real workloads to open weights is just a matter of time. openrouter shows how, swapping a frontier model for an open one cuts the per-token cost ten to fifty times. the split isn’t a prediction. it’s a config change.

the material.

• epoch ai — open models lag closed by four months ↗ read

• yash patil — open weights are a leading indicator ↗ read

• openrouter — how to get the lowest-cost inference ↗ read

things to notice:

• the frontier’s whole premium is a four-month head start. for good-enough work, four months is nothing.

• the gap is slightly widening (three months last fall, four now). bears see a moat (a four-month head start wins a winner-take-all race), bulls see a rounding error (nothing, for good-enough work). which read is right depends on the game you’re playing.

sit with these. then take them to claude.

• what do you still pay a premium vendor for that an open model now does at a tenth, trust or inertia?

• if the model trends toward free, where does the money go, and who was just reselling it?

• when frontier intelligence is a four-month-old free download, what’s the new scarce thing?

iv. why does cheaper make the bill go up?

cheap wakes up demand that was asleep.

so why did the bill explode if the price collapsed? jevons paradox. when something gets cheaper, hidden demand wakes up and you use far more, not less. packy’s version is the clearest: cheap intelligence means more lawsuits, not fewer lawyers. better design demanded everywhere. that’s why cost-per-answer falling and the total bill exploding are the same fact.

the material.

• packy mccormick — intelligence superabundance ↗ read

things to notice:

• efficiency creates demand, it doesn’t satisfy it. the most counterintuitive thing in the room.

• jevons has a precondition: the cheaper thing must unlock real, productive demand. that’s the bet, not a law.

sit with these. then take them to claude.

• the last time something got cheap for you, did you save the money, or use way more of it?

• if cheaper ai means more total spend, who wins on volume even as prices fall?

• what doesn’t exist yet only because the tokens still cost too much?

v. if it’s getting cheaper, why aren’t the returns showing up?

not yet, and the two ends are pulling apart.

the steelman. marcus stacks it: ninety-five percent of pilots show no return, uber can’t find proportional gains, “if enough companies admit the same, the story breaks.” then the same tunguz who told you the price would explode downward adds the sharp part: the frontier is getting more expensive. google tripling prices, opus at five dollars in and twenty-five out, because the cheap years were a subsidy that ends when margins matter. the two ends pull apart.

the material.

• gary marcus — if enough companies report the same, the bubble pops ↗ read

• tomasz tunguz — the unsustainable subsidy ↗ read

things to notice:

• “the price fell” and “the economics work” are different claims. most noise wins the first and pretends it won the second.

• watch the divergence: open gets cheaper, frontier gets dearer. two curves, opposite directions.

sit with these. then take them to claude.

• of the ai you pay for, how much would you miss if it vanished, what would you defend with your own card?

• if frontier prices rise while open prices fall, which side is each company you follow on?

• when the subsidy ends, who’s left, the makers, the picks-sellers, or the ones built on top?

vi. what’s still scarce when intelligence is free?

you are. the human in the loop.

end on the deeper question, and the hopeful one. if intelligence gets nearly free, “too expensive” stops mattering, and what stays scarce starts to. imas and trammell work it out: when machines make most things abundant, human involvement becomes the scarce resource, the work where a person in the loop is the value. azeem points at the upside, cheap intelligence is an entrepreneurship boom, because a mediocre ai lawyer beats none.

the material.

• dwarkesh × alex imas + phil trammell — what’s scarce after agi ↗ listen

• azeem azhar — the ai boom is becoming an entrepreneurship boom ↗ read

things to notice:

• when the input gets free, value runs to what stays scarce: taste, the human in the loop, the uncopyable.

• the same cheapness that threatens the analyst is what lets one person start the firm. the floor drops for you too.

sit with these. then take them to claude.

• what part of your own work would be worth nothing if a machine did it, where you are the point?

• if intelligence is the free input, what scarce thing gets repriced up, and who owns it?

• what would you build the day a team of ai workers costs less than lunch?

cheap is the easy part. it’s coming for everything, on a schedule, whether we’re ready or not. the part still open is what survives it. and this whole thing is named for the answer: when the intelligence is free, you’re what’s left. the taste, the trust, the read that can’t be downloaded. so don’t ask how smart the model is. ask which side each company builds on, and what it makes scarce on purpose. then go make yourself the thing that can’t get cheap.

— brylan.

tea’s steeping.

slow afternoon, nobody’s rushing.

kick your shoes off.

you ever notice the difference between the friend who picks up right where you left off. and the one you re-explain your whole life to every time.

same coffee. same conversation. but one of them is building on you. the other starts from zero. with one, the years add up. with the other, they just pass.

i couldn’t stop thinking about it this week. it’s the same line running under the whole ai argument right now.

what i’m calling it.

the thaw.

for years the smartest models on earth were frozen. trained once, shipped, done. same answers forever. you could talk to one every day for a year and on day 366 it knew you exactly as well as it did on day one. which is to say. not at all.

but watch how it’s been moving. first these things just answered questions. a smarter search box. then they did tasks. you hand over the whole job to an agent and it comes back done. now the people building them talk about handing it a responsibility. it runs in a loop, watches a problem, owns the outcome.

and here’s the part underneath. you can’t hand a responsibility to something that forgets. a thing that owns an outcome has to learn from what just happened. take the correction, fold it in, come back sharper. that’s the shift. frozen, to learning.

but learning is a slippery word, so let me be clear. memory is notes about you. learning is a changed mind. a friend can keep a list of everything you’ve ever said and still be the same person every time you talk. what matters is whether the mind underneath actually moved.

so the line is. does it learn from you, or does it forget you. does the mind underneath get sharper the more you touch it. or does it just keep more notes.

where it shows up.

once you’ve got the line, the whole world sorts into frozen and thawing. most things still forget you. the form that asks for your address the fourth time. the company that treats every call like you’ve never called. a vending machine. same input, same output, your thousandth visit identical to your first. that’s most of what we use. the whole bet of this moment is melting it.

it’s been melting in stages, and most people never clocked it. first, databases. the machine kept what you typed in, nothing more. then retrieval. it could look things up before answering, pull in your own docs. then memory, the kind claude and chatgpt have now, carrying what matters from chat to chat so you stop re-introducing yourself. each step feels like learning. but it’s the notes getting better, not the mind. the brain underneath never moves.

the real thaw is the next step, the one the whole field is reaching for. the mind itself changing from your use. the bet a wave of startups is making now is the signal everyone else throws away. every time you accept what it gives you, retry, or fix it yourself. that’s the gold. the answer matters less than what you did with it. it’s hard for a real reason. teach one of these things something new and it can lose what it already knew. nobody’s solved that yet. but every lab is pointed at it and billions are flowing in to crack it. you can feel how close it is and how not-here-yet it is at once.

this is still mostly on screens. picture it with a body. waymo already runs one driver across its whole fleet, but today that driver is frozen between updates. trained in a batch, shipped, the same until the next version. now picture the thaw reaching it. a car that learns a tricky left the moment it takes it, and the whole fleet has it by morning. that’s what’s coming for all of robotics. the arm on the line. the thing folding laundry in somebody’s kitchen. each one getting better the second it does the work, and teaching every other one while it does. things with bodies are just slower to thaw.

how i actually use it.

point the line at anything you put real time or money on.

is this compounding on me, or am i starting over.

• your money: this is the lens for the whole ai market right now. the openai, anthropic, and google races. the ipos everyone’s waiting on. don't ask how smart the model is. ask who owns the notes. the coding tool that gets sharper every time someone accepts or tosses what it wrote is compounding on millions of people at once, and it keeps that record. the wrapper renting somebody else's model keeps nothing. the day the big platform ships that feature for free, the wrapper's gone. the loop, and the notes underneath it, are not. that's the thing nobody can buy off a shelf.

• your people: who do you leave a little more known. and who keeps you re-introducing yourself. not every bond has to compound. some people you just love. but know which is which, because the closest ones are usually the ones who kept building on you.

• yourself: are you learning from your own corrections. or shipping the same mistake on day 366 you shipped on day one. most people aren’t frozen because they can’t change. they’re frozen because they never fold the lesson back in.

here’s where it leaves me.

it used to take months to train the next model. a new brain once or twice a year, frozen in between. now the whole field is rowing toward a model that learns every time someone uses it. and someone’s using it every second. data already moves as fast as we do. if intelligence starts moving that fast too, the next version stops being a thing that ships. it becomes a thing that never stops thawing. that’s not here yet. but it’s the direction. so when you look at any of it, hold one question. is this building the thing that thaws, or renting one that’s already frozen. i’d rather watch it move than wake up on day 366 surprised it did.

— brylan.

coffee's hot.

the world's still waking up.

come sit.

spacex is about to be the biggest ipo that has ever happened.

a target around 1.9T. bigger than facebook. bigger than alibaba. bigger than aramco.

that’s interesting.

this is more interesting.

look at the red part. 2020 to 2025. 70 years of human space history, and the last five took the whole curve and bent it.

spacex flew 165 falcon 9 missions in 2025, more than the rest of the world combined, and the dominant share of all mass to orbit per jonathan mcdowell’s tracker.

while we were watching the gpt-5 demos, that happened.

and once i saw it, the rest of the s-1 started reading differently.

what is actually happening.

so what is spacex now, actually? because it’s not just a rocket company anymore.

first they learned how to move things. then they learned how to build things. now they’re learning how to power and compute things.

that’s the shape. five layers. they’re in every one.

move things (launch). a falcon 9 has two parts that matter. the booster (the big bottom that pushes everything off the ground) and the second stage (the smaller one that takes the payload the rest of the way up). spacex figured out how to catch the booster, refurbish it, fly it again. the rest of the world is still throwing theirs away. that’s how one company ended up doing 80% of the mass. and most of that mass is their own. starlink launching starlink. that’s not a launch business winning a market. that’s a manufacturing machine feeding itself.

build things (satellites). they don’t make satellites the way satellites used to be made. they make them on a line. nearly 10,000 starlinks in orbit, per jonathan mcdowell’s tracker. that’s not a space program. that’s an automotive factory pointed at the sky.

power things (terrestrial compute). somewhere along the way, the rocket company started pouring concrete. colossus + colossus II = 1.0 gw of compute online in memphis. yes, really. (this is also the xai side of the house. the two merged in feb 2026. spacex didn’t build the ai layer. it bought it. but the manufacturing logic is what makes the purchase make sense: you don’t buy a compute business, you buy a factory you already know how to run.)

silicon. most people see ai and think software. eventually every ai story becomes a factory story. that was the whole point of last week’s syllabus. the s-1 names a thing called terafab. a “chip manufacturing initiative with tesla and intel” whose long-term goal is one terawatt of compute hardware per year. that’s the biggest leap in the whole stack. tesla and intel making a terawatt of chips a year is a stated ambition, not a fab that exists, and it’s the line i’d bet against soonest. but even as an ambition it tells you what the company thinks it is: no longer a launch business. and from the same filing, a quieter line that load-bears the whole stack: “the future of ai will be determined by the control of the physical stack.”

compute in space (orbital). google announced project suncatcher in november 2025. solar-powered tpu constellations in low earth orbit. planet labs builds the satellites; google is reportedly in talks with spacex for the launches. prototype mission slated for early 2027. this is the bet at the top of the stack. solar-powered compute that doesn’t need a grid or a cooling tower, because space already is one. this is where last week’s syllabus on the physical layer beneath the ai story and gavin baker’s framing land: cities are already banning new data centers, so the next ones have to leave the planet.

okay but how does this all hang together financially?

launch sells at internal marginal cost to starlink. starlink prints the cash. 11.4B in revenue in 2025, 49.8% year-over-year growth, and roughly 38.8% segment margins per the s-1. the cash funds the ai bet (which lost billions and ate the majority of capex).

that’s the architecture. spacex is a vertical manufacturing loop that uses launch as a cheap toll booth. the rocket isn’t the product. it’s the toll. and every dollar it collects gets reinvested up the stack.

what i noticed.

once you see the stack, you start noticing things in the s-1 that didn’t make sense before.

this week i kept staring at three numbers. they’re all the same number, really. they all say spacex is a manufacturing system pretending to be other things.

fingerprint one. colossus.

colossus I took 122 days to build. colossus II took 91. that's 25% faster, build over build. the same compounding curve they ran on falcon 9 boosters, now on data centers. the company that learned how to shorten rocket cycles is shortening building cycles.

fingerprint two. the wedge.

look at how flat the rest of the world is. then look at the red. spacex deployed 3,180 starlinks in 2025 alone. more than the rest of the world combined. while everyone's debating chip allocation, spacex turned a satellite line into a car factory. the wedge is a manufacturing curve, not a launch curve.

fingerprint three. the cash engine.

starlink subs more than doubled in 2025, from 4.4M to 8.9M, and hit 10.3M by end of q1 2026. revenue tripled in two years: 3.9B → 7.6B → 11.4B, still growing 49.8% at 11B scale. that doesn’t happen at this run-rate very often. this is what shortening production cycles does to economics. cost falls. base widens. cash compounds. the curve isn’t done.

every one of these is the rocket logic applied somewhere else. the rocket was practice.

how i’m seeing it.

so what do you actually do with this. here’s how i’m seeing it.

one sentence that holds the whole thing together: the thing that could break in the next twelve months is the thing that compounds for thirty years if it doesn’t.

let me show you both sides.

the break. 71% of the valuation sits on a segment that lost 6.4B last year.

meritech’s s-1 breakdown (by alex clayton + austin wang) backs into the rumored 1.7T number, and ai does 71% of the work to get there. they’re explicit: they’re not justifying the number, they’re showing what it takes. most of the company sits on the segment that lost the most money. that could go in twelve months if ai demand softens or chip allocation pulls back.

the compound. semianalysis broke down what the cost stack actually looks like today. space datacenter compute: $8.64 per gpu-hour. terrestrial: $2.37. that’s a 3.6x premium.

the gap isn’t engineering. it isn’t power. it isn’t operating cost. it’s mostly one line. datacenter capital cost. $6.29 vs $0.36 per gpu-hour. that’s launches.

which means the cost curve mostly tracks one variable: what falcon and starship cost per kilogram to orbit. cooling, radiation, and replacement cycles aren’t free up there either, but launch dominates the gap and it’s the one falling fastest. every dollar off launch comes off the orbital compute curve. it’s one model, not a law. parity around 2035 in the musk scenario, 2039 in the semianalysis base case. either way the bear case is already a ramp. the direction is the part i’d bet on.

the shift in between.

which raises the next question: does the ai bet actually pencil? if it does, the whole loop runs. if it doesn’t, the toll booth is still a toll booth. just less interesting.

the honest answer: not yet, and not for a while. and here’s the part that keeps me patient about that. the binding constraint right now isn’t power. it’s chips. semianalysis calls it the great ai silicon shortage. ai demand eats roughly 60% of tsmc’s leading-edge output in 2026, around 86% in 2027. and that constraint is universal: it doesn’t care whether the datacenter is in memphis or in orbit. space doesn’t solve the chip problem. it solves the power problem. but only in the world where ai demand is so enormous it’s blown past everything we can build on the ground. that’s why this is a thirty-year bet, not a twelve-month one. it’s also why terafab exists: the only way to beat a chip constraint is to make more chips.

the 30-year thesis in one line: launch costs commoditize → spacex owns the cheapest path to orbit → orbital infrastructure becomes a real market. everything else is texture.

some lingering thoughts.

and here’s what would make me change my mind. not as disclaimer. as a dashboard.

the worst thing i could do is fall in love with the story. so here’s what i keep turning over.

the ai buildout’s capex still hasn’t shown up in revenue. david cahn at sequoia has tracked the gap since 2023. 125B, then 600B, then trillions. it widens every update.

if the labs never print fast enough and the whole stack reprices, who’s left holding spacex’s ai segment when it reprices too?

then the rest of what i keep turning over:

• starship is the load-bearing assumption under the entire cost curve. three of the last eight tests were ruds. rapid unscheduled disassembly, what spacex calls “it blew up.” they’re learning fast. but what if fast still isn’t fast enough, and rapid reuse slips three years? does the whole orbital thesis just sit and wait?

• colossus went from 122 days to 91. but what if that pace was memphis. just the water, the power, the permits lining up. not a repeatable machine? how much of the manufacturing fingerprint is the company, and how much is one good location?

and the one that’s really been on my mind:

the entire ai segment is two rental contracts. and it carries 71% of the valuation. anthropic pays 1.25B/month for the whole output of colossus 1. then three days ago, google signed too. 920M/month for gemini enterprise. ~26B a year, combined. (yes, the same google that’s in talks to launch suncatcher. it shows up on both ends of the stack now.)

two of the best-funded companies on earth. that sounds like diversification. it isn’t. both leases end in 2029. both tenants are ai labs. which means both are exposed to the exact capex vs revenue gap from the chart above. if that gap forces a reprice, it doesn’t take one customer. it takes the whole segment, at the same time. so what is the ai segment actually worth if both leases come up for renewal into a market that’s decided ai compute was overbuilt?

if all of it breaks at once, the rocket is still cheap. it just stops being the factory equipment for anything bigger. still a launch monopoly. just not 1.9T.

i don’t know which way it goes. nobody does. but i know what i’m watching, and i know what would change my mind.

what i’m sitting with this week: i thought i was looking at a rocket company. i was actually looking at a manufacturing machine that kept expanding into adjacent layers. that reframe is the only thing that made the s-1 stop feeling weird.

if launch costs commoditize the way semianalysis thinks they will, who owns the orbit?

— brylan.

no agenda.

oat latte’s hot.

sit down. laptop open.

last month i hit “you’ve reached your usage limit” on claude in the middle of a workday. not surprising. it’d been happening more. but this time i sat with the popup instead of refreshing. somebody, somewhere, decided i needed to stop. and the reason wasn’t software.

a few weeks earlier i’d been listening to jensen huang on dwarkesh. he said the thing that flipped it for me. the bottleneck on the next decade of ai isn’t an algorithm. it’s plumbers and electricians. shortage of trades. transformers on backorder. substations waiting on permits. dario amodei said it from the demand side. anthropic planned for 2 or 3x growth. the business did 80x and has since blown past a $47 billion run rate. “the constraint becomes physical infrastructure.” altman said it shorter. openai’s chips were overheating from how many people were using chatgpt.

the rate-limit popup and the plumber are the same story.

by plumber i mean the whole physical layer beneath the model. the electrician. the hvac tech. the substation. the chip. the cooling tower. the water pipe. everything the model is quietly sitting on top of.

the lesson.

claude. chatgpt. gemini. the new ai app that shipped this week. each of them is in a different fight with the chips, data centers, power, and water they need. some can still absorb demand. some are already constrained.

the wall you hit on a tuesday afternoon is infrastructure talking back.

the syllabus.

five things to sit with. in this order if you can.

each one's a link. open it, sit with it, come back.

i. usv, the myth of the infrastructure phase. ↗ read

the spine of this whole piece.

dani grant and nick grossman push back on the idea that we move through clean “infrastructure phases” followed by “app phases.” they argue infra and apps pull each other forward in loops. apps create demand. demand pulls new infra. new infra unlocks new apps.

things to notice:

• once you see the loop, you stop asking “are we in the infra phase or the app phase.” you start asking which side is pulling harder right now.

• the rate-limit popup is the application side pulling new infrastructure into place in real time.

sit with these. then take them to claude.

• which ai product can do something this year it literally couldn’t last year, and what got built underneath to make that possible?

• which piece of infrastructure is getting ordered right now because an app blew past everyone’s expectations?

• what other industry runs on this same loop and nobody calls it that?

ii. jensen huang on dwarkesh. ↗ listen

the framing.

jensen describes nvidia as “the in-between.” electricity goes in. tokens come out. one sentence reorganizes the whole map of the ai business. it’s an industrial company, not just a software one.

things to notice:

• listen for the part about elon having more nvidia chips than openai and anthropic combined.

• pay attention to how often he talks about energy, throughput, and factories instead of “ai magic.”

sit with these. then take them to claude.

• nvidia turns electricity into tokens. what’s the next “in-between” company nobody’s named yet?

• if elon is sitting on the most compute, what does he have to ship to turn it into revenue?

• where’s the chokepoint inside the inference economy itself?

iii. dario amodei on dwarkesh. ↗ listen

the math.

dario said: “if my revenue is not $1 trillion, if it’s even $800 billion, there’s no force on earth that could stop me from going bankrupt if i buy that much compute.” every weekly cap on claude is downstream of that one sentence.

things to notice:

• listen for what he thinks anthropic’s revenue actually has to look like by 2028 to make the spend safe.

• the existential risk for anthropic isn’t a bad model. it’s the bill.

sit with these. then take them to claude.

• if dario is right, what does using claude or chatgpt feel like in 2030?

• who’s writing the check that keeps anthropic alive in 2028?

• does the winner end up being the smartest model, or the one that figured out how to pay for itself?

iv. ben thompson, agents over bubbles. ↗ read

the economics.

thompson argues an “agent” is a model plus a harness. the moat lives in the harness. the infrastructure, the integrations, the runtime. not the weights. if he’s right, the whole “best model wins” frame is wrong.

things to notice:

• the valuable layer may not be the model. it’s the system wrapped around the model.

• every successful harness drives more compute demand underneath it. harness winners and chip winners aren’t opposed. they pull each other.

sit with these. then take them to claude.

• which boring office task that took you hours or days just became automatable because of opus 4.7 agent capabilities?

• which workflow nobody’s wrapped yet could be worth more than the model running underneath?

• which job is one capability away from being done by software, not a person?

v. gavin baker on orbital data centers. ↗ listen

the frontier.

baker on what happens when “more power, more land, more cooling” stops being something you can permit on earth. cities are already banning data centers. the response is to leave the planet.

things to notice:

• the picture in your head is probably wrong. it’s not pentagon-sized facilities in orbit. it’s racks.

• the deeper idea is that compute is becoming geographically mobile. it doesn’t have to live in a building.

sit with these. then take them to claude.

• if a rack can run in orbit, what’s the next strange place compute ends up? underwater. in your car. on a balloon. what does each one unlock?

• what else are we about to launch into orbit because we ran out of room down here?

• what becomes possible when compute is no longer tied to a building?

so. look upstream. what else has to grow to support the ai you’re using?

— brylan.

nobody’s in a rush in here today.

trust yourself.

pull up a seat.

you ever tell yourself monday you’re locking back in. five days at the gym. early nights. the book on your nightstand. by wednesday you’ve been twice and stayed up too late twice. you almost text a friend “i’m so back.”

but you don’t send it. because you’ve been here before.

wanting to believe yourself. but you’re humble enough to know you need to see it first.

what i'm calling it.

proof, not promises.

it's a read i started running on myself first.

you don’t get to call yourself changed because you wanted to be. you don’t get to call it a new chapter on a monday. proof is the version of you that keeps showing up. over and over. when nobody’s watching. when nothing’s on the line.

it’s the gap between what somebody says and what actually shows up after they say it. promises live at the front of the gap. proof lives at the back. most of life happens in the middle. the wait. the part where you find out if the words had legs.

so you let the gap sit. you don’t move your life around what they said until the proof catches up.

where it shows up.

people. companies. whole industries. the gap stays the same. you just have to be patient enough to let it sit.

every company in tech is out here promising ai is about to replace entire workflows. but the receipts are still thin. revenue hasn’t moved. margins haven’t moved. most people still work basically the same way they did a year ago.

nvidia is the other side of the same coin. proof has been showing up every quarter for two years. i kept waiting for the slowdown and it hasn’t come. when the receipts keep showing up. you update. you don’t get to keep your old read just because it felt smart.

spacex is the one that snuck up on me. data centers in space sounded like a slide deck for years. but in november somebody flew an nvidia chip on a rocket. ran a model up there. it worked. the promise of orbital compute quietly became a receipt. nobody noticed.

how i actually use it.

i started doing this because the world got loud and i didn’t trust how much of the loudness was getting inside me.

now i keep three things in the front of my mind. what got promised. what would prove it. when i’d know. that’s it. write it down somewhere only you see.

once the receipt is written down somewhere, the story gets quieter. you’re not arguing with the news. you’re not pre-deciding. you’re just waiting for something specific to either show up or not show up. both answers are information.

the practice doesn’t tell you what’s true. it tells you what you don’t know yet. it gives you a place to put the not-knowing so it doesn’t run your week.

that’s the whole thing. that’s the play.

— brylan.

what’s playing: lock in. — gunna, key glock, babyface ray.

i’ve been wrestling with figma for weeks. filings. transcripts. building a model just to argue with my own model. the fun has been the wrestling itself.

what i keep coming back to is the gap. between what the market is saying about figma and what figma is actually doing. between the picture and the felt-thing.

i’ve felt this one before. dinner went fine. the movie went fine. the picture from the outside was clean. the felt-thing underneath was something else. you can talk it out. it doesn’t change the past.

that’s the shape of this page. that’s also the shape of figma right now. the market is debating whether ai kills figma. the real question is whether figma is still where software gets made.

how this kind of business works.

figma sells a file. the rest is dressing.

the file is a multiplayer browser canvas where designers, developers, pms, and now ai agents meet to build software. before figma, designers worked in sketch, exported pngs, and emailed them around. figma did to design what google docs did to word. it made the file the meeting place. once the meeting place is yours, every adjacent workflow becomes a feature you bolt on and charge for.

they charge by seat, sliced by role. designer. developer. collaborator. content. same playbook as office charging readers, writers, and editors different prices for the same documents. when a big customer adds engineers and marketers over time, the bill grows even if the seat-price doesn’t.

here’s the wild part. the customers who paid figma more than ten grand last year are paying thirty-nine percent more today. same customers. one year later. thirty-nine percent more. that’s after everyone who quit, every downgrade, every account that got rationalized. the bill grows because the file grows into more of the building. the engine works. that’s the part to remember.

the story being told.

i watched it drop. read the substack posts. felt the panic on twitter.

the story is that ai is killing figma.

it’s loud. plaintiff law firms circling. anthropic’s head of design left to ship a figma competitor. she was figma’s own director, the one who built figjam and slides. google launched two design products in march. anthropic launched claude design in april. every drop has had a date and a story attached. that’s what makes it so easy to tell.

the case has a name. ai bundling. anthropic and google and lovable give designers most of what figma gives them, for free, inside the same window they already use for code. small users leave first. enterprise follows. the spending-more pattern breaks. figma becomes adobe at half the price.

that’s the market. that’s what’s loud right now.

what the market is expecting.

figma went public august 1st, 2025. it ripped to a hundred and forty-two ninety-two within weeks. nine months later it touched sixteen sixty-nine. today it sits at twenty-two fifty-two. eighty-three percent off the high.

to price a stock down that hard, the market needs a story.

i built the model to figure out which story. sat with it for a week. it kept asking me what i actually believed about the next three years.

here’s the one it’s telling.

designers don’t have to draw anymore. type “build me a settings page that matches our app” into claude or cursor or v0, and the ai ships working software in seconds. no mockup. no figma file. no handoff. jenny wen said it on lenny’s podcast in march. the design process is basically dead. she’s anthropic’s head of design now. former figma director. figma was the venue for the old way.

if she’s right, designers shrink as a role. fewer seats sold. customers who used to spend more start spending less. ai is expensive to run, so the dollars figma keeps cost more to deliver. investors decide each dollar is worth less. figma ends up looking like adobe at half the price.

that’s the bet. at twenty-two dollars and fifty-two cents, the market is pricing it. figma’s cash grows sixteen percent a year for the next decade. not the thirty-five they’re currently telling wall street. claude design and canva take real customers. all of it has to be roughly right for the price to be roughly right.

i built the model. screenshot below. file linked at the bottom.

two more things. eleven percent short interest. about six million shares betting figma falls further. anything good prints, they have to cover into a thin tape.

and a stock moves on two engines: the dollars the company makes, and what investors will pay per dollar. they can move opposite ways. goldman cut their price target to thirty dollars in april and raised the dollars they expect figma to make. the damage was investors paying less per dollar. not figma making fewer.

let me show you something.

i read the print twice. once for what it said. once for what it didn't.

ten days ago, figma reported its first quarter. revenue up forty-six percent from a year ago. the customers-spending-more thing? they went from spending thirty-six percent more to thirty-nine percent more. accelerating, not slowing. after paying to deliver the product, figma still keeps seventy-nine and a half cents of every dollar. after paying the bills, they keep sixteen cents. they’d told wall street to expect nine.

three out of four of the bigger customers used figma’s ai features in the most recent week. that’s people leaning in. the opposite of leaving.

the price implies sixteen percent long-term growth. the company printed forty-six. that’s the gap. that’s the whole trade.

what i think.

who is figma, actually? not who figma is to the substack writer who tried claude design over a weekend. who figma is to the seven out of ten customers buying more than one of their products. who figma is to the customers spending thirty-nine percent more this year than last. a company where designers are now a minority of the people inside the building. where engineers are the fastest growing group. where every new ai feature is another way to charge, not another way to lose.

the right comparison isn’t adobe. it’s coordination. github. slack. atlassian. companies that got more valuable as the work they coordinated got harder, not less. if engineers become the largest seat type inside figma, figma stops being “design software.” it starts being where teams of humans and agents meet to make software.

that’s a different company than the one being shorted.

history has two shapes for moments like this. in one, the market is right early. kodak. blockbuster. nokia. numbers look fine for a while. then they don’t. in the other, the market is wrong. microsoft in 2014. adobe in 2011. disney through streaming. the story said extinction. the numbers said adapting. the stock eventually followed the numbers.

the evidence so far points toward pattern two. a figma director put thirty-six and a half million dollars of his own money into the stock in february. cathie wood added eight million in march, after google launched. the bigger customers are spending more, not less.

dylan field is the other tell. defensive language is what you’d expect from management watching the funnel break. instead, figma raised expectations through the panic. quietly started training their own ai model in august 2024, twenty-two months before config. spent around two hundred million buying weavy in october. people running for the exits don’t do that.

here’s the deeper reason incumbents sometimes survive these moments. ai changes how a single artifact gets generated. it does not change the work of getting a hundred people aligned on which artifact to ship. governance. approvals. design systems as infrastructure. multiplayer state. enterprise memory. auditability. the file isn’t valuable because it’s hard to make a mockup. it’s valuable because it’s where the company decides what to build. coordination is the moat that doesn’t go away just because generation got cheap.

history rhymes here. the cloud commoditized compute, and orchestration layers became valuable. open-source commoditized software, and the workflow layers became valuable. ai might commoditize interfaces. if it does, what becomes valuable is whatever does governance, alignment, and coordination across humans and the agents now in the market with them. figma is closer to that surface than anyone in the market is currently pricing.

but at this stage in 1996, microsoft looked like pattern one too. you don’t know which one you’re in until you’re two or three years past it. my confidence the case for figma is defensible. that a smart person could argue it without lying. sixty out of a hundred. my confidence figma actually wins over the next two and a half years. forty.

the market is shorting a 2027 obituary that 2026 keeps refusing to print.

the edge.

i'll admit it. i bought the consensus story for a minute. it sounds smart. it's been wrong for a year.

the case the market is repeating goes like this. ai makes figma’s customers spend less. it has been wrong for four straight quarters. they’re spending more. the prediction is dead. the price still acts like it’s alive.

the real risk is harder to see and harder to tweet. it’s not that customers spend less. it’s that fewer new customers show up at the door. canva picks off small teams before they try figma. claude design picks off the people who’d have mocked something up over a weekend. figma becomes a company that sells only to big enterprises and grows slower. by the time you see it in the numbers everyone watches, it’s already been true for years.

the market is debating whether ai kills figma. that fight is settled in figma’s favor for the next four to six quarters. the real question is whether the top of the funnel is already silently broken. we won’t see it until contract renewals roll through in 2027 and 2028.

that’s where i diverge.

what would confirm this.

i have these two dates circled.

here’s what i’m actually watching. three drivers and four kill conditions. all of them resolve at the next print or before.

drivers that prove the hypothesis right:

1. the bigger customers keep spending more. the net retention number stays above one hundred and thirty percent. it’s a hundred and thirty-nine today.

2. revenue stays up at least thirty-eight percent from a year ago. it was forty-six this quarter.

3. figma keeps at least seventy-nine cents of every dollar after delivering the product. it was seventy-nine and a half this quarter.

4. after the bills, they keep at least ten cents. they kept sixteen this quarter.

5. the ai features start showing up as real money on the income statement. quantified, not hand-waved.

6. dylan ships figma’s own ai model at config. trained on the billions of files figma owns. with a real number showing it beats claude or openai on design tasks.

kill conditions. any one of these and i exit:

1. the net retention number prints below a hundred and thirty in any quarter.

2. a named enterprise customer publicly cuts its figma seats and cites a claude design or stitch.

3. two quarters in a row of fewer than five hundred new ten-thousand-dollar customers.

4. dylan field departs or sells a material insider stake.

two dates carry the test. config 2026, june 23rd through 25th, thirty days out. q2 2026 earnings, august 18th, eighty-six days out. config is the strategy read. q2 is the numbers read. together they tell me whether to ride the option or close the position.

this isn’t a forever bet. it’s an eighty-six-day bet with defined exits.

where i could be wrong.

the part i keep losing sleep over isn’t in the financials. it’s the workflow.

old way: idea → figma → review → engineering → code → deploy.

new way: idea → prompt → generated app → iterate live → deploy.

if the new way wins, the question of where collaboration happens stops having figma as its default answer.

jenny wen isn’t alone. lovable does a hundred thousand new projects a day. cursor runs in eighteen percent of professional developers at work. v0 has six million developers. these tools aren’t on most people’s radar yet. time-to-idea is collapsing in a place i cannot fully see.

the honest version. i can model the financials. i can’t model the shift in how product gets built with the same conviction. the real risk isn’t ai killing figma. it’s a new shape of work where figma isn’t the venue. big company contracts run one to three years. the spending-more pattern can hold for another four to six quarters before this shows up. then it shows up all at once.

the picture and the felt-thing can stay split for a long time before anyone knows which one was real.

some lingering thoughts.

• if i’m right about the prints but wrong about the workflow, how long does the trade have before the architecture question catches up to it?

• where does coordination actually live in an ai-native software stack? if it’s not the figma file, what is it?

• what does jenny wen know from inside anthropic that the figma prints haven’t shown yet?

• am i sympathetic to figma because i use it every day?

the market is loud about the wrong thing. the case it’s repeating has been wrong for a year. the case i’m actually worried about is invisible right now. the numbers say one company. the stock says another. the next ninety days will tell us which one we’ve been looking at.

the work, for any of us, on any company we’re trying to read, is seeing them for who they are. not who the market says they are. not who we wanted them to be. who they actually are when you sit with them long enough. figma’s real question isn’t whether ai kills it. it’s whether it’s still where software gets made. the answer to that one won’t come from august.

i haven’t seen what i need to see yet. but i’m sitting closer.

and i’ll tell you straight. i bought a long-dated call on figma earlier this month. thirteen hundred and twenty dollars. skin in the game. not a recommendation, not a heavy bet. the size that says i think this, but i could be wrong, and i want to be honest about both. writing a page like this without telling you would be hiding the ball.

the option runs to january 2028. the bet is ninety days. if config and q2 deliver and the stock pops on a short cover, the trade was the squeeze and i sell. if it doesn’t pop, the option lets me hold for the architecture question to resolve. which it won’t by august. the mismatch is on purpose. but i should name it so neither of us forgets which one i’m actually trading.

— brylan.

the sources: every figma-specific number on this page is grounded in figma’s own investor materials. the 2025 10-K. the Q1 2026 10-Q. the Q1 2026 earnings call and presentation. the Q4 2025 earnings call. all on figma’s investor relations page and on sec.gov. the rest. jenny wen on lenny rachitsky’s podcast. claude design on anthropic’s site. lovable, cursor, and v0 numbers from sacra. all of it is public. all of it is one search away.

disclosure: long fig via a jan 2028 $10 call. opened may 8th, 2026 at $20.04 underlying for $13.20 in premium. one contract. this is not investment advice.

what’s playing: the night after the argument got quiet. — big k.r.i.t., quasar, leisure dept.

we were arguing about the war in the groupchat. the usual thing. who’s right, what’s really happening, what it means.

i laid out my read. the K.

how where you land on it shapes everything you see after that. a few people pushed back. it got heated in the way groupchats do.

fast, then quiet.

and then he shared his portfolio.

no words. just the screenshot. and something in the conversation ended without anyone saying it was over.

i sat with that for a while tonight. trying to figure out what it meant. and then, in a different conversation it hit me. he manages hundreds of millions of dollars. if i were him, i don’t know if i would have stayed in the conversation either.

that’s the thing i can’t stop thinking about.

what is actually happening.

there’s a war. and depending on where you sit. financially, politically, geographically. it looks completely different. not like people are seeing different details. like they’re observing different realities.

this is the K.

economists have had a name for it since 2008. one path climbs. one descends. the S&P has climbed over one hundred and thirty percent since march 2020. the top one percent hold nearly thirty-two percent of the nation’s total wealth. the highest since records began in 1989. the top ten percent drive half of all consumer spending. half.

this week, one man’s net worth crossed eight hundred billion dollars. more than the GDP of switzerland. more than the bottom fifty percent of americans combined. in december 2024 he was worth four hundred billion. he added the second four hundred billion in fourteen months. fourteen months.

on the other side of that K.

she works two jobs. nursing. fills her tank twice a week to get there. gas was under three dollars before february twenty-eighth. it crossed four dollars last week. she’s doing the math every time she pulls up to the pump. groceries were already up thirty percent since 2020. now fertilizer prices are spiking because the strait of hormuz is closed, which means the food she’s already stretching to afford will cost more by summer. she’s not watching the war on a bloomberg terminal. she’s watching it at the register.

same war. different country.

same economy. different country.

and how you land on the K isn’t just about your politics. it’s about your position. what you own. what you owe. what a different outcome costs you. your financial position shapes your observation apparatus — the way a lens shapes what light reaches your eye. you’re not choosing to see things differently. you’re structurally positioned to.

the friend who walked away wasn’t wrong. he just couldn’t see what i was pointing at from where he was standing. and i couldn’t see what he already knew from where i was standing. the argument wasn’t just heated. it was already over before it started.

what i noticed.

upton sinclair said it cleaner than anyone. it is difficult to get a man to understand something when his salary depends upon his not understanding it. he wasn’t talking about stupidity. he was talking about structure. the incentive reshapes the perception before you even notice it happening.

that’s what the portfolio screenshot was. not deflection. his answer. the most honest thing he could say in that moment.

i am here. this is what i see from here. your argument doesn’t reach this position.

the research backs this. paul piff at UC Berkeley has spent twenty years studying what capital does to people. as wealth increases, empathy goes down. entitlement goes up. self-interest feels more justified. and the disturbing part — the distortion starts before the money is even real. in one study, fake monopoly money was enough to change behavior within minutes. that’s not a character flaw. that’s a mechanism. and it operates on everyone — including you.

what i’m starting to understand is that this isn’t a bias you correct for. it’s a pressure that builds with capital. the position doesn’t just make certain truths expensive to hold. at some point, it makes them impossible to reach.

ambiguity is a temporary privilege.

the honest answer is i stayed in the conversation because i’m not fully on either side yet. i can still see both. and somewhere underneath the argument i knew that. the window doesn’t stay open.

how i’m seeing it.

paul tudor jones is one of the most aggressive traders alive. he also runs robin hood, one of the largest poverty-fighting foundations in the country. he doesn’t resolve the tension between those two things. he holds it. his line: i am a capitalist. and i believe capitalism has to evolve. not abandoning the game. playing it differently.

chamath’s frame is the one i keep returning to. standing at stanford, he said it plainly: “money drives the world for better or for worse. you have a moral imperative to get it and then use it to make a difference.” and then, almost as an aside: “there are about 150 people who control most money in the world and they don’t give a shit about what you think about their worldview. so if i can accumulate capital, why shouldn’t i deploy it against mine?”

get the money. then create new rules.

but the part of that sequence nobody talks about: more capital doesn’t automatically produce more generosity. in some cases, it does the opposite. the sequence only works if the person who arrives at the money is still the same person who set the intention. that’s the variable nobody accounts for.

howard marks draws a distinction i think about now. the best investors don’t suppress the feeling. they separate it from the action. the emotion exists in one room. the trade lives in another. that’s not detachment. that’s compartmentalization with integrity. knowing which room you’re in. not pretending the other one doesn’t exist.

different people solve for the pressure in different ways. none of them remove it.

some lingering thoughts.

if i’m honest: if i were managing hundreds of millions of dollars, would i have stayed in the groupchat? i want to say yes. but i don’t know. and the not knowing is the part that matters.

james baldwin wrote that nothing can be changed until it is faced. what i keep sitting with is the inverse. every conversation you exit is a data point you stop collecting. not just a social thing. an epistemological one. the conversations you stop having narrow what you’re able to see. quietly. gradually. before you notice.

the higher your position, the more expensive it becomes to hear things that contradict it. so you hear less of them. and over time, your reality narrows — without you noticing.

it’s less about character than people want to admit. and more about position than they’re comfortable facing.

what am i willing to be wrong about? that i can still see both sides. that the window is still open. that i’m writing this from somewhere neutral. i’m not. i’m writing this from a position too. one comfortable enough to aestheticize the tension, articulate the mechanism, and still not name what it’s already hiding from me.

that’s the question i can’t shake: what are you already not seeing from where you currently stand?

she’s still at the register. doing the math. that’s not a metaphor. that’s just tuesday.

the chart: when the market recovers fast and wages lag behind, ownership decides who wins.

the watch: trump says the economy is strong. voters disagree.

i’m still in the conversation. i don’t know for how long. i’m watching what that costs me. and what it costs to leave.

— b.

what’s playing: slow burn awareness — mereba, odeal, tyla, barry can't swim.

saturday afternoon. airport. gate c14. flight back to chicago.

i’m the one not looking at my screen.

everyone else is. laptops open. headphones in. fingers moving. doing things. processing. navigating menus, clicking through tabs, composing emails, filling out forms. the ordinary machinery of a workday that followed them to the terminal.

i’m sitting here watching them and wondering. not judging. wondering. what they’re afraid of. what they’re hoping for. how they’re using ai today. whether they feel the ground shifting underneath all that clicking.

earlier this week, on st. paddy’s day, i showed up to a hackathon in sweats. two hours to build something with a team i’d never met. i could do everything. i knew how to break the problem into pieces. give each piece to the right machine. get something back. i tried to delegate. my teammates stalled. not because they weren’t smart. they were. but they sat in front of the same interface i was using and came up empty. they were asking the machine what to do. i was telling it what i needed.

we didn’t get to pitch. i wanted the room to see what we’d built. how. so they could learn.

the outcome-thinker didn’t get the outcome. that’s worth holding too.

small room. big room. same question.

i remember when they taught us how to type.

what is actually happening.

the word “computer” first appeared in 1613. it meant a person.

someone employed to perform calculations. following fixed rules. processing information with precision. no authority to deviate from the instructions given. for three hundred years, a computer was a human being doing what we now describe machines doing.

you can look them up. katherine johnson. who computed orbital trajectories by hand for nasa’s early missions. the women at harvard’s observatory who catalogued hundreds of thousands of stars. not seen as scientists, but as computers. the women of langley. hired because the job required patience and precision. not because anyone thought they were building careers.

then in 1945, the machines arrived. the word migrated from the person to the machine. the humans who programmed it were recruited from the same pool they were replacing. they knew the work. they just had to teach the machine to do it.

and then we decided everyone needed to learn to talk to the machine.

home row. ctrl+c. file → save as. thirty years of mandatory curriculum built around a single premise. to be useful in the modern economy, you must learn to operate the machine.

what we were really teaching was translation. how to encode human intent into machine language. computer literacy wasn’t about computers. it was about learning to be the bridge between what you wanted and what the machine could do.

thirty years of humans learning to be used by machines well.

three things changed that.

the first was chatgpt. november 2022. made the machine readable. but it got things wrong and required you to catch the mistakes. you had to bring the agency. the skill was still yours.

the second was reasoning. september 2024. the machine started checking itself. the burden of verification started lifting. you were still in the loop. but the loop was getting smaller.

the third is what most people haven’t registered yet. computer use.

sometime in late 2025, ai learned to use a computer the way you do. open a browser. navigate to a site. read what’s on the screen. click the button. fill out the form. send the email. not by being told each step. by being given an outcome and figuring out the path itself.

anthropic moved first. claude cowork. a desktop agent that lives on your computer, accesses your files, controls your browser, executes tasks autonomously. you describe what needs to happen. it makes a plan and works through it. it felt less like a tool and more like leaving messages for someone who actually does the work.

openclaw followed. an independent developer built an open source agent that runs on your own hardware. accessible through whatsapp, telegram, imessage. and can control your browser, manage your files, and keep working while you’re away from your desk. in three weeks it became the most downloaded open source project in github history, surpassing thirty years of linux adoption. jensen called it “probably the single most important release of software, probably ever.”

then perplexity launched computer. nineteen specialized frontier models orchestrated automatically. you describe an outcome. the system routes each piece to the right model and runs them in parallel. for hours or months. all while you do other things.

three products. one direction. the machine does the clicking now.

people don’t just want smarter answers. they want something that acts.

the computer learned to use itself.

most people are already living in the gap between what the economy demands and what they can do.

the oecd found that only five percent of adults score at the highest level of computer proficiency. a quarter can’t use one at all. and ninety-two percent of jobs require digital skills.

computer use either closes that gap or makes it permanent.

what i noticed.

this week i built seventeen skills. sets of instructions that tell the machine what to do when i show up with what i want. the machine already knows its job.

for example, i onboarded partners in atlanta. this requires creating the partner in the system, setting up dashboards, drafting the emails, looking up contacts. a task that requires twelve steps. and back and forth across multiple tabs. this is just one example of where i provided the context. the machine handled the execution.

you’re the context. the machine is doing the clicking.

the gap i saw at the hackathon wasn’t technical. my teammates weren’t less intelligent. they were thinking in tasks. asking the machine what to do. i was thinking in outcomes. telling it what i needed. that difference, right now, is the whole ballgame.

but here’s what i keep sitting with.

three out of four workers who try ai tools abandon them mid-task. not because the tool failed. because they couldn’t tell if it succeeded. the adoption gap isn’t capability. it’s trust. and trust requires being able to check the work.

i think about my brother. a lot, this week.

he’s smart. genuinely smart. but the digital world was never built for how he thinks. he struggles with the basics. not because he can’t learn, but because the basics were always arbitrary. copy-paste. right-click versus left-click. where the file went after you downloaded it. why the same button does different things in different apps. he’d call me and i could hear the frustration. not “i’m stupid” frustration, but “why does this have to be this hard” frustration. and he was right to ask. it shouldn’t have been.

for thirty years, we built interfaces that demanded you memorize their language. ctrl+c on windows, command+c on mac. file → save as. navigate to the folder. name it something you’ll remember. millions of small translations between what you meant and what the machine required. my brother isn’t bad at computers. he just never accepted the terms of the translation. and nobody told him that was a valid response.

ai doesn’t automatically fix that for him. in some ways it makes it worse. the expectation shifts. if the computer can use itself, the question becomes, can you at least confirm it did it right? computer use hands you leverage you didn’t have before. but only if you know what good looks like on the other side. my brother has never been given the tools to know what good looks like. not because he can’t. because no one designed that bridge for him.

the gap doesn’t close for him. it widens.

and he’s not alone. anthropic interviewed eighty-one thousand people across a hundred and fifty-nine countries. what people asked for most wasn’t speed. it wasn’t efficiency. it was presence. something that actually understands their situation. the people who need ai most aren’t asking for a faster tool. they’re asking for a kinder interface. one that meets them where they are.

watching the agent work this week, i realized what digital work actually is. not thinking. navigating. clicking. formatting. switching apps. hundreds of micro-decisions per task. the agent absorbs all of them.

but here’s the question i haven’t resolved. if reasoning models already check themselves and orchestration systems verify across multiple models, then verification itself might be the next thing that gets automated. my brother’s gap might not be permanent. it might be an interface failure that the next generation of tools actually solves. maybe the right agent doesn’t ask him to verify. it asks him what he wants and confirms it got there. maybe the bridge isn’t literacy anymore. maybe it’s design.

i don’t know. i want that to be true. i’m not sure it will be. but i’m not willing to assume my brother’s gap is forever when the interfaces that created it are the very thing being replaced.

how i’m seeing it.

cowork made a specific bet. the value is in knowing your context. your files. your folders. your workflows. built into your machine. perplexity made the opposite bet. the value is in knowing which model to use. nineteen models, routed automatically, each best in class at a specific kind of work. they’re playing the orchestra, not the instrument.

both bets are live. but the direction underneath both is the same. the model is becoming invisible. the interface that deploys it is what people experience.

openclaw proved the demand was real before both of them. three weeks to surpass linux’s thirty-year adoption curve. jensen at gtc: “every company in the world today needs to have an openclaw strategy. this is the new computer.” and his trillion dollar revenue forecast is arithmetic, not ambition. by nvidia’s estimate, every agentic task consumes around a thousand times more tokens than a single prompt, and the number of people handing outcomes to machines instead of clicking themselves is just getting started. we’re at an inflection where “the last prompt was queries. this prompt is actions. do something for me.”

the market is still largely pricing the model as the moat. but last january, two models held ninety percent of enterprise usage. by december, no single model held more than twenty-three percent. meta paid two billion for manus. not for a foundation model, for the layer that tells the computer what to do. apple built the platform everyone is running this on. they just haven’t claimed what it’s becoming yet.

the value isn’t in building the smartest model anymore. it’s in building the thing that knows when to use which one.

when the economy tightens, the jobs that go first are built on execution. the clicking, the navigating, the following of fixed rules. the same jobs thirty years of typing classes prepared people for. with hormuz closed and costs climbing, companies aren’t waiting for the comfortable moment. the war created the cover. computer use was already waiting. the timing isn’t coincidental. it’s compressive.

what survives is knowing what you want. and knowing how to tell if you got it. the typing class taught you to operate the machine. computer use means the machine operates itself. that’s what the typing class was for. and for most people, it turns out it was preparation for something that no longer needs them to show up.

some lingering thoughts.

• the typing class was mandatory. someone decided the skill mattered enough to require. nobody is requiring the next one. what does it mean that the most important literacy of our time is opt-in?

• when your manager can hand an outcome to an agent instead of delegating it to you, what’s left of the entry-level job? where do people learn judgment if they never get to do the work that builds it?

• if the value moves from knowing how to use the tool to knowing what you want from it. what happens to the people who were never taught to want clearly? who were taught to follow instructions, not give them?

• my brother’s boss sees him as slow. what happens when the boss gets an agent that’s fast and decides the gap between them is no longer worth closing?

• being early and being wrong feel the same from the inside. what am i not seeing?

the read: agents over bubbles — ben thompson — the clearest framework i found this week for why computer use changes the economics of everything above the GPU.

the product: perplexity computer — don’t just read about it. understand how it works. this is what the shift looks like when it’s a product.

the rabbit hole: letter 8 — global hub theory — kp — went against every narrative i was holding. left me asking where i need to revisit my assumptions.

the question i can’t shake: what happens to the people on both sides of the computer literacy divide when the computer starts using itself?

the people who built careers on the translation. using excel. powerpoint. navigating the browser. formatting the spreadsheet. sending that follow-up. the machine can do all of that now. the people who never learned. the computer was supposed to be their way up. now it’s doing the clicking.

both sides lose something when the computer starts using itself. but maybe the same shift that makes the skill obsolete also makes the interface kinder. maybe the bridge my brother never got is the one being built right now.

what grows in that space is still being decided.

they taught us to use computers. nobody taught us what to do when computers use themselves.

that’s the part we figure out now.

— b.

what's playing: what's going on. — marvin gaye, sam cooke, le'andria johnson, curtis mayfield.

jennifer hudson sang “a change is gonna come” at jesse jackson’s funeral this week.

if you know the song’s origin, you know what that moment was carrying. sam cooke didn’t write it because he was certain. he wrote it because bob dylan, a white folk singer from minnesota, had written “blowin’ in the wind” and cooke heard it and felt something he couldn’t name. something close to shame. that someone else had seen his people’s pain more clearly than he had. before he was ready. before the room was ready.

the song took him a year to finish. not because he didn’t know what he wanted to say. because he did. and saying it out loud in a world that wasn’t ready meant carrying it alone until it wasn’t.

then i watched jesse jackson’s face at obama’s election. tears on his cheeks. not grief. not surprise. recognition. a man who’d spent forty years in that room. carrying something the room wasn’t ready for. watching it land somewhere real. finally.

i thought about that face a lot this week.

i’ve been in a lot of arguments.

what is actually happening.

on february twenty-eighth the US and israel launched operation epic fury. over a thousand targets hit in the first twenty-four hours. b-2 stealth bombers. bunker-busting bombs. the supreme leader killed. the most concentrated display of american air power in decades.

the official reasons kept changing. trump said the goal was eliminating imminent threats. and urged iranians to take back their country. rubio said the US had to move first because it knew israel was about to strike, which would have triggered iran to hit american forces. hegseth called the mission laser-focused. destroy the missiles, sink the navy, no nuclear weapon. when trump’s words sounded like regime change, hegseth pushed back. “this is not a so-called regime change war. but the regime sure did change.” by day four, a senior democrat on the senate intelligence committee said he’d watched the stated goals shift four or five times.

three spokespeople. four or five stated goals in four days. most people in my world grabbed one explanation and ran with it. and the arguments started.

here’s what most people miss.

venezuela and iran aren’t separate moves. they’re the same move.

china is the world’s largest importer of oil. roughly half of everything it imports flows through the strait of hormuz. the narrow waterway between iran and the arabian peninsula connecting the persian gulf to the rest of the world. when epic fury started and iran began retaliating, that strait effectively closed. tanker traffic dropped between seventy and ninety-four percent. shipping insurance suspended. oil moving toward a hundred dollars a barrel.

china keeps emergency reserves for exactly this kind of disruption. by current estimates, three to four months of runway before those reserves run low.

the trump-xi summit is in four weeks.

venezuela’s oil got sanctioned earlier this year. iran’s oil is now disrupted. those are two of china’s biggest energy sources. cut off at the same time. xi walks into april’s summit watching his country’s energy clock run down, sitting across from the man who started it.

that’s leverage. and it’s leverage that took something to create.

here’s the uncomfortable part.

to defend against iranian drones and missiles, the US relies on interceptor systems. expensive rocket-powered missiles, made by lockheed martin and raytheon, that shoot down incoming threats mid-air. they’re the shield. and shields run out.

iran has fired more than two thousand drones and five hundred ballistic missiles since epic fury began. the US has been intercepting them across five countries simultaneously. israel, the UAE, kuwait, bahrain, and qatar. each one drawing from the same finite stockpile.

the administration argues they’ve been going after the source. hegseth said the US is hunting and killing iran’s launchers. by march fifth, ballistic missile attacks were down ninety percent from day one. drone attacks down eighty-three percent. the case is that destroying launchers solves the math. fewer things to intercept means the shield lasts longer.

that argument is real. but it rests on whether they can find launchers faster than iran can relocate them, and whether the stockpile holds in the meantime. each iranian drone costs roughly thirty-five thousand dollars to build. each US intercept costs between five hundred thousand and four million. the math is uncomfortable regardless of whose numbers you use.

the US military’s central command has reportedly told planners internally this war could run through september.

seven months. and china is watching every move.

what i noticed.

some analysts are saying this is catastrophic for china. that xi’s energy story broke. that gulf states pivot toward the US. that china has no good options. compelling. might be right. but it rests entirely on iran being quickly and decisively finished.

the iranian foreign minister’s interviews don’t sound like a regime in collapse. calm. organized. not asking for a ceasefire. that’s a regime that has decided survival is the strategy. absorb the strikes. keep firing. wait out the political clock. the US midterms are twenty months away. iran doesn’t need a military victory. they need a negotiated exit before american public opinion runs out. hezbollah ran this playbook against israel in 2006. they didn’t win. they didn’t lose. the war ended. they were still standing.

the part that doesn’t get said plainly, the US proved it can project force at this scale, and in doing so revealed exactly what it costs. the leverage on china and the vulnerability to china landed in the same week.

the GCC open letter barely made the front pages. the saudi-led bloc of gulf states, the region the US is supposedly protecting, publicly rebuked trump. estimated forty to sixty-five billion in direct costs, two hundred ten billion with spillover. the alliance that keeps global oil priced in dollars is quieter and more fragile than it looks.

russia didn’t fire a shot. sanctions eased. oil prices rose. western attention split. putin came out ahead and almost nobody led with that.

how i’m seeing it.

china was already under pressure before epic fury started.

their real estate market, which makes up a larger share of the economy than US housing did before 2008, is still unwinding. major developers defaulted. millions of apartments sold to buyers who haven’t seen them completed. overall debt is near three hundred percent of what the economy produces in a year. youth unemployment near seventeen percent. twelve million college graduates entering the job market this year. the official growth target is the lowest since 1991. xi has been prioritizing political control over the structural reform that might actually fix it.

a disrupted energy supply isn’t just bad news in that context. it’s a pressure point arriving at exactly the wrong moment.

but here’s what i kept sitting with all week. what’s happening in the middle east isn’t just about the middle east.

taiwan is a self-governing island of twenty-three million people. china has claimed it for decades. xi has called its reunification a historical inevitability. the only thing that has consistently kept china from moving on it is the belief that the US would intervene. and that the US military could win if it did.

both of those beliefs just got complicated.

the USS abraham lincoln carrier strike group, previously stationed in east asia specifically to deter china, was redeployed to the middle east to support epic fury. analysts at the stimson center wrote in foreign affairs this week that US indifference to taiwan, combined with watching interceptor stockpiles deplete in real time, could lead china to conclude a window has opened. the asia times put it plainly, the interceptor shortage isn’t just a problem in iran. it’s a strategic inflection point for taiwan.

the honest counter is that the offensive display is its own deterrent. hitting a thousand targets in twenty-four hours, killing a supreme leader, coordinating across five countries. that’s a message. the sword matters, not just the shield. china has to weigh whether american willingness to act at that scale changes the calculus even with a lower stockpile. that argument is real.

but the carrier group that’s no longer in the pacific is not theoretical.

my group chat got hostile the same way rooms always get hostile when someone is carrying something the room isn’t ready for. not because they had better information. because it was easier to defend what they already believed than to sit with both things being true at the same time.

the leverage thesis and the vulnerability thesis aren’t opposites. they’re the same event seen from two angles. the US created pressure on china. and revealed exactly how deep the magazine goes in the same move, in the same week. that’s the thing the chat couldn’t hold. not one truth or the other. both at once.

xi shows up to that table with his energy clock running down. his economy already strained. and a window over taiwan that serious analysts are now saying may never be more open. that’s not a negotiating position. that’s an existential calculation.

the grand strategy might be right. the leverage might work. china might take a deal. and all of that can be true while the cost of creating it reshapes the next decade of deterrence in ways nobody is fully pricing yet.

can the US end this fast enough to matter or has it already spent the credibility it needed to deter the next war?

i came into this week confused. i don’t feel confused now.

i feel something else.

some lingering thoughts.

• are launches in iran down because the launchers are gone or because iran is conserving?

• hezbollah survived because its leadership was still standing. does decapitation change the endurance math?

• if the window over taiwan is opening, when does china decide it sees it?

• being early and being wrong feel the same from the inside. how do you know which one you’re in?

• what am i willing to be wrong about? the grand strategy might be too elegant. blunt force with convenient side effects isn’t the same thing as sequenced design.

the watch: strait of hormuz tanker traffic. if it normalizes within thirty days, the leverage thesis holds. if it stays suppressed, the vulnerability thesis starts to dominate. that’s the clock.

the read: zineb riboua’s "the iran question is all about china" and "china is scrambling." she makes the strongest case that epic fury broke xi's story. read both, then sit with the counter.

the question i can’t shake: if the US wins this and it still costs more than anyone priced, what does that tell china about the next one?

obama stood at jesse’s funeral and said, don’t put your head down and wait for the storm to pass.

he remembered watching jesse debate on a tiny black-and-white TV with rabbit ears, and thinking. he owned that stage. he wasn’t a pretender. he belonged there. and if he belonged there, maybe so do i.

jesse spent forty years in rooms that weren’t ready. carrying a clear read. saying send me. not i know, but send me. the loneliness wasn’t permanent. it was the price of being early.

a change is gonna come. not a prediction. a posture.

keep hope alive.

— b.

what’s playing: late night thinking session — mikecol, quasar, lion babe, gold link.

spent my birthday in orlando with my mom and sister. we watched the spacex rocket launch. i don’t have words for it yet. just that i’ll never forget it.

got back to chicago. the lake was frozen solid. something about seeing it from the plane. still, quiet, holding everything underneath.

then there was the tesla earnings call. elon mentioned terafab. robotaxi. amazing abundance. the stock’s at four eighteen. watching something take off is one thing. evaluating whether this one will is another.

and dating. i spent time with two women since i landed.

dinner wednesday night. she’s been showing up differently. on time. asking questions. transparent about where she stands. the hug after felt different. the three-hour facetime felt like home. but i’ve been here before. i know how this can go. so i’m paying attention. skeptical but open. wanting to believe, but needing to see it first.

then friday night. electric. we caught a vibe. but the signs were there. small comments, inconsiderate moments, things that didn’t add up. i noticed. didn’t act. by the end of the night it came out sharper than it should have. we won’t see each other again. she’s not wrong for how she moves. she’s just not my person. i don’t need to teach reciprocity. i just need to notice faster and leave earlier.

same week. one where the change might be real. one where i knew better and stayed anyway.

so what does this have to do with tesla?

how this kind of business works.

tesla makes money three ways today:

• cars. about seventy billion in revenue, but shrinking. down ten percent last year.

• energy. nearly thirteen billion, almost doubled in two years. backlog strong.

• services. supercharging, insurance, parts. another twelve and a half billion.

but the interesting part isn’t where the money comes from. it’s how the pieces connect.

they own the stack. cars generate data. data trains the software. software enables autonomy. autonomy unlocks robotaxi. robotaxi funds the next thing. energy powers all of it.

the third thing isn’t a business yet. it’s the bet.

robotaxi. optimus. physical ai. if it works, they’re not selling cars or batteries. they’re selling labor hours and transport miles. think airbnb for cars. you let your tesla work while you sleep. recurring revenue at software margins on physical infrastructure they already own.

they just sunset the upfront fsd purchase. subscription only now. that’s a signal. they want recurring revenue, not one-time payments. the business model is evolving.

and now they’re going deeper.

elon announced a “terafab” on the earnings call. a us-based chip factory for logic, memory, and packaging. he’s spending his saturdays on the ai5 chip design. they’re building lithium refineries and cathode facilities. they want to make their own solar cells.

model s and x are being discontinued. fremont is becoming an optimus factory. one million humanoid robots a year is the target.

the mission got updated. tesla is now chasing “amazing abundance.”

they’re not asking you to see them as a car company anymore. they’re asking you to see them as infrastructure.

the story being told.

every few years the story shifts. electric vehicles. full self-driving. robotaxi. optimus. each time the current thing gets challenged, a new future thing justifies the valuation.

that pattern makes me cautious. it’s like watching someone promise they’ve changed, then promise again, then promise again.

but here’s what’s different this time. it’s not just narrative.

gross profit held despite revenue decline. they’re doing something on the cost side. r&d spiked sixty percent. they’re putting money where the words are. and they have the balance sheet to fund this themselves.

forty-four billion in cash. around fifteen billion in annual operating cash flow. and they just guided to over twenty billion in capex this year. six new factories, ai compute, infrastructure expansion.

this isn’t a company hoping narrative will fund its runway. they’re writing the check themselves.

and they’re burning the ships. discontinuing s and x to make room for optimus. converting factories. going all in on the identity shift.

i respect conviction. i’ve learned the hard way that half-measures don’t work.

but the story still has to become real. and that’s where i’m watching closely.

what the market is expecting.

at four hundred thirty dollars a share, what are you actually paying for?

the reverse dcf breaks it down:

• current business value. sixteen to forty-three dollars per share (seven to ten percent of the stock price).

• ai/robotics optionality. about four hundred per share (ninety-three percent of the price).

the market believes tesla will grow revenue six to seven times from here. ninety-five billion to five fifty or six sixty-five billion. that’s twenty-plus percent annual growth for ten years to fifteen years.

but auto is shrinking. energy is growing fast, but it’s only thirteen percent of revenue. so ai and robotics have to not just grow. they have to more than offset the decline in the core.

implied margins at maturity. fifteen percent or higher. current is four point six.

this isn’t a “show me” valuation. this is a “believe me” valuation.

let me show you something.

the reverse dcf splits the price into two pieces.

piece one. the current business. using an ebit multiple, about sixteen dollars a share. using a revenue multiple, maybe forty-three. midpoint around thirty. either way not much.

piece two. everything else. four hundred dollars betting on robotaxi, optimus, and ai infrastructure that hasn’t arrived yet. ninety-three percent of the stock price is optionality.

but here’s the caveat. i’m modeling something i can’t fully see. they’re spending twenty billion on capex next year. when do we get disclosure on cost per ride, margin per ride, utilization rates? until then, part of this is faith.

if robotaxi delays. if the story breaks. the stock reverts to car company multiples. fifty to a hundred dollars. seventy percent drawdown.

the math says you’re paying full price for a future that hasn’t arrived yet.

this isn’t a mispriced stock. it’s a binary bet.

either they become the physical ai platform and this goes to eight hundred, or the narrative breaks and it reverts to one thirty.

what i noticed.

• automotive revenue dropped, but it tracked with deliveries. no pricing collapse. just softer demand. that’s different from a brand problem.

• gross profit held despite the revenue decline. they’re doing something on the cost side. discipline showing up when it matters.

• r&d spiked sixty percent. that’s not maintenance spend. that’s them putting money where the narrative is. ai, robotaxi, optimus.

• they sunset the upfront fsd purchase. subscription only now. they want recurring revenue, not one-time payments. but it might also mean the eight thousand dollar option wasn’t converting.

• elon said austin robotaxis are running “without chase cars.” reporting from days before said chase cars were still present. that gap. between what’s said and what’s verified is exactly what i’m watching.

• energy had a record quarter. one point one billion in gross profit. this segment is becoming material while auto shrinks. and it’s not dependent on the robotaxi narrative.

how i’m seeing it.

i believe in the direction.

production is the moat, not software. i keep coming back to tesla as an infrastructure company. service centers, charging, manufacturing. the question isn’t whether fsd beats waymo on a benchmark. it’s who can actually scale hardware.

commodity dynamics protect against brand risk. if it’s cheaper, people take it. brand matters at the margin, not the decision. that’s a bet that price wins. and tesla is getting better at cost.

the owner fleet model makes sense to me. airbnb for cars. you let your tesla work while you sleep. the unit economics are intuitive if utilization is high enough.

but i’m skeptical on the price.

the story is ahead of the physical reality. digital ai is in millions of hands today. tesla’s physical ai is still in “trust us, it’s coming” territory.

same pattern i’ve seen before. someone says they’ve changed. the words are right. but i need to see it sustained before i believe it.

where i diverge from the bulls. the gap between narrative and physical reality is wider than it looks. surface signals aren’t enough for me.

where i diverge from the bears. the tam might be much bigger than anyone models. and in the ai age, narrative can sustain valuations longer than fundamentals suggest. plus they have the cash to keep building regardless.

the edge is patience. wait for the gap to close. either price comes down or reality catches up.

what would confirm this.

• unsupervised robotaxis without conflicting reports

• published safety data comparable to waymo

• robotaxi revenue in the financials

• unit economics disclosed

• cybercab actually shipping in april

• optimus actually in factories

proof, not promises. same standard i hold for anyone trying to show me they’ve changed.

if the price pulls back to one-fifty or two hundred, the risk-reward changes materially.

where i could be wrong.

time might be the weapon i’m underestimating. belief attracts capital, capital buys time, time delivers results. if they execute, four eighteen looks cheap in hindsight. waiting for proof might mean buying at six hundred.

and the reflexivity loop is real. tesla’s high stock price isn’t just a scorecard. it’s a currency. it lets them hire the best engineers with stock comp. the “bubble” valuation might be the fuel that builds the thing that justifies the valuation.

i could be wrong about the moat. i keep saying production, not software. but what if software is actually the bottleneck? what if waymo’s approach ends up winning because regulators demand it?

china is the risk i can’t model away. elon admitted it on the call. byd is eating the low end. xiaomi is moving fast. that’s where the brand premium erodes first. and geopolitically, that revenue concentration is a vulnerability.

some lingering thoughts.

• what if waiting for proof is the wrong frame? what if belief is the catalyst that makes the proof possible?

• why is elon spending saturdays on chip design? what does that tell you about where he thinks the bottleneck is?

• how long can the balance sheet buy them before the market demands results?

• where could i get hurt even if my thesis is right? timing, position size, opportunity cost?

• what if patience is just a story i tell myself to avoid risk?

the history lesson: a documentary on hitler. sounds unrelated. but i've been thinking about the last time geopolitical tensions were this high. how narratives build. how belief compounds. how people saw the signs and stayed anyway.

the chart: tesla has the miles. waymo has the rides. the gap.

the read: bill gurley on tam. the ride he thought would take thirty minutes took five. the market wasn’t what he modeled. it was ten times bigger.

the question i can’t shake: what does it cost to need proof before you believe?

still thinking about that launch. the sound of it. watching something actually take off.

i don’t know if this is that. but i’m not looking away.

— b.

what’s playing: floors turning to marble — christian kuria, tom misch, darius, j cole.

i turned thirty-one this week.

the morning of, i woke up to a text from j. he wrote: “god is going to honor your sacrifices and restore what you’ve lost. actually, restoration is already in progress. i can tell by the way you speak.”

i sat with that. he sensed something in me before i had words for it.

the night before, i woke up at 4:53am crying. still dreaming about someone who isn’t here anymore. questions with no answers, spinning in circles. then the tears came. i deleted old wallpapers of us. no chantilly cake this year. no balloons. no note. just me.

but j saw something else. motion. something already happening that i hadn’t named yet.

this week i didn’t consume much content. but i sat and noticed. davos happened. the world economic forum. bloomberg events. the adults in the room gathering. and from a quick view, something was aligning with how i’ve been seeing the world.

a shift in trump’s tone. canada’s tone shift. carney saying the “rules-based order” is a lie we told ourselves. ken griffin and larry fink diverging. claude on a rocket ship. dario as the modern day oppenheimer.

the world is already moving. most people just haven’t felt it yet.

what is actually happening.

think about the bank you drive past. the local branch. the one your parents used. maybe you still have an account there. ask yourself. why does it still exist?

for decades, the answer was simple. banks borrow cheap and lend expensive. they pay you half a percent on your savings, charge someone else seven percent on a loan, and keep the spread. that was the business model. that was the reason to exist.

but here’s what most people miss.

stablecoins now pay four to five percent. backed by treasuries. same safety. better yield. the coinbase ceo said it plainly this week. banks feel threatened. six trillion dollars could leave traditional banks if stablecoins pay competitive interest.

six trillion. that’s not market share. that’s the whole reason the bank exists.

and this isn’t happening in a vacuum. trust is leaving the system.

foreign central banks are buying gold, not treasuries. the us is tariffing the people who buy our debt. bessent’s “america first” is becoming “america alone.” carney at davos said middle powers now need “variable geometry”. ad hoc alliances because the old ones don’t hold.

the ground is shifting. the old pipes are cracking.

and then there’s ai.

nathan lambert wrote something this week that landed: “being good at using ai agents is a better moat than working hard. agents push the humans up the org chart.”

2026 is the year of the agent. claude just shipped cowork. agents that can run for hours, managing files, drafting, researching. satya calls them “token factories.” openai, google, anthropic, meta. they’re all shipping agents that can work for days without stopping.

the skill that made you valuable. the thing you trained for. the grunt work you cut your teeth on. is becoming ai’s baseline.

and people feel this. the fear. the anxiety. the vague sense of “am i on the wrong side of this?” that’s not paranoia. that’s the body knowing before the world catches up.

here’s the common thread: intermediaries are losing their reason to exist.

banks extract rent from deposits. workers extract rent from scarce skills. both are losing the reason they exist at all.

what i noticed.

• the market sees chronic stress. business as usual with friction. vix at sixteen. banks still open. jobs still posted. the headlines feel loud, but the prices stay calm.

• but the body knows.

• the vague unease when you check your savings rate. the job that feels different even if the title is the same. the sense that something is moving beneath the surface.

• at davos, the adults in the room were worried. blackrock, citadel, lagarde. drawing parallels to the 1920s. isolationism, tech boom, debt. the “great moderation” is over.

• some people are already repositioning. moving to the new pipes. learning to manage agents. holding assets outside the old system.

• most are still standing still. walking past the bank like it will always be there.

how i’m seeing it.

i don’t think the old system is dying slowly. i think it’s losing its reason to exist. that’s different. slower deaths leave time to adapt. losing your reason to exist is structural. it doesn’t come back.

near-term, the infrastructure is being built. the genius act passed. nyse is working on tokenization. twenty-four-seven trading, instant settlement, stablecoin compatibility. agents are shipping. the new pipes are filling up.

medium-term, the world catches up. the first bank reports deposit flight. the first company announces workforce reduction from agents. the market reprices.

long-term, the world looks different. some of what we rely on today won’t be there.

but here’s what j saw in me that i’m starting to see in the world. maybe the shift isn’t loss. maybe it’s clearing. making room.

the dot-com boom ran on debt. the ai boom runs on balance sheets. that difference matters. this isn’t speculation. it’s infrastructure. and infrastructure tends to stick.

things exist in duality. the same forces that make intermediaries obsolete are building something new. the question isn’t whether the shift is happening. the question is what survives it.

some lingering thoughts.

• what if the body is early, but not right?

• when money changes, who gets left behind? what happens to the coffeeshop on the block?

• if agents do what juniors do, where do juniors go?

• is this intuition or just hope?

• what survives the shift?

the chart: stablecoin market cap. three hundred five billion and climbing. the new pipes filling up.

the read: nathan lambert — “being good at using ai agents is a better moat than working hard.”

my view: themes, ideas, and catalyst.

the question i can’t shake: what stays when the rest falls away?

j saw it before i did. restoration is already in progress.

the world is moving. the body knows.

i’m not waiting for permission anymore. i’m already in motion.

— b.

what’s playing: dad’s r&b and morning energy. — kenny lattimore, mary j. blige, larry june, smooth grooves.

a lot happened this week. and somewhere in the middle of it, i realized i’m different now.

i went back to my old stomping grounds. walked the campus. met with students. faculty. saw old faces. they made it feel like home. but i wasn’t the same person who used to walk those paths. something had shifted in me.

i traded futures with my cousins. we hadn’t had a moment like that in a minute. not as kids, as adults. talking positions, risk, life. it hit different.

i bought myself flowers. just because.

i had a hard 1:1. the kind that sits in your chest after. the kind you want to run from. i didn’t run.

the loneliness has been loud. yearning for a hug. a kiss. something. i don’t have answers. but i’m sitting with it now. not running. that’s new.

and then pastor mike’s sermon landed all of it:

“i’m not giving up. i’m growing up.”

“surrender is not passive. it’s disciplined.”

“outcomes belong to god. reactions belong to us.”

“god doesn’t force. he forms you.” see becoming happens through proximity that shapes you, process that stretches you, and practice that seals you.

i’ve been holding those words. turning them over. that’s what this week was. formation. not chaos. shaping.

and then i looked at the world and saw the same thing happening.

rearranging.

what is actually happening.

trust is leaving the system. you can feel it even if you can’t name it.

this week the department of justice served the federal reserve with a criminal investigation. the official reason was cost overruns on a building renovation.

the way i’m reading it, this is about pressure.

trump needs lower interest rates. now.

here’s what most people miss. the US owes more than our entire economy produces in a year. nine trillion dollars of that debt is coming due in 2026. most of it was issued when rates were near zero. now it has to be refinanced at four or five percent. that’s like refinancing your mortgage at double the rate.

interest payments on the national debt already exceed what we spend on defense.

trump needs lower rates. but if he pressures the fed too hard, investors lose trust, rates go up, and the problem gets worse.

powell said it plainly: “this unprecedented action should be seen in the broader context of the administration’s threats.”

and regular people are feeling the squeeze too.

real wages were flat in december. groceries jumped 0.7%, the biggest spike in over three years. rent keeps climbing.

american workers are now taking home their smallest share of the economy since 1947. that’s almost eighty years. the pie is growing, but workers are getting a smaller slice.

from 2000 to 2022, median household income rose seventy-eight percent. home prices rose one hundred seventy percent.

flat wages plus unaffordable housing means most people can’t move up. if you don’t already own a home or have family wealth, breaking in is nearly impossible. your starting position determines your ceiling.

trump is responding.

he announced plans to cap credit card interest rates at ten percent and ban large investors from buying single-family homes. americans owe 1.23 trillion dollars in credit card debt at around twenty percent interest. cutting that in half could save tens of billions a year.

banks are pushing back. their stocks dropped. JPMorgan hinted at lawsuits. their argument is if rates are capped, they’ll stop lending to people with bad credit. and the perks you’re used to like travel points, cash back, sign-up bonuses could shrink or disappear.

whether these policies work is another question. but the fact that a president is intervening this directly tells you how broken the math has become.

and here’s what most people aren’t paying attention to.

the old system locks people out of wealth. the new system is trying to build a way back in.

right now, when you send money through a bank, it takes days to settle. the bank earns interest on your deposits. you get almost nothing.

that’s starting to change. stablecoins are digital currencies backed by dollars. and they’re growing fast. the companies that issue them earn interest on the reserves. and now there’s a fight in congress about whether they can pass that interest to you.

the ceo of coinbase pulled support for a bill this week because it would ban that. banks are lobbying hard to stop it. the ceo of bank of america warned that if stablecoins can pay interest, six trillion dollars could leave traditional banks.

six trillion times three and a half percent is two hundred ten billion dollars a year. that could flow to regular people instead of bank profits.

the old pipes are cracking. new pipes are being built. most people are still arguing about the faucets.

what i noticed.

• countries and companies are buying gold, silver, and copper. not because they’re scared. because they’re positioning for a future that looks different than the past.

• gold is protection. when you don’t trust governments or currencies, you buy gold. central banks have been stockpiling it for years. china and russia especially.

• silver is protection plus demand. it’s in everything being built. solar panels, electric vehicles, electronics. not just fear. real use.

• copper is in everything. every wire, motor, data center, EV, drone. AI infrastructure alone is adding massive demand.

• the financial cracking and the physical scarcity are the same story. countries see it. companies see it. the question is whether regular people see it in time.

how i’m seeing it.

i don’t know where this goes. the old anchors are slipping.

near term, the doj investigation looks like leverage, not prosecution. the real target is rates. but the pressure itself is the signal. if the fed can be pushed around, the dollar’s credibility takes a hit.

medium term, whoever controls the new pipes wins. stablecoins, digital money, the ability to move value without middlemen. and whoever controls the physical stuff like copper, rare earths, energy has leverage in a world where trust in paper promises is fading.

i learned something this week about what i can and can’t control.

i made three thousand dollars trading copper and gold. good setups. patience. discipline. then i got greedy on a gold trade. no stop loss. no protection. things turned against me fast. six thousand two hundred dollars gone in minutes.

the thesis was right. the execution was wrong.

i wanted to control the outcome. i couldn’t. all i could control was my reaction. my process. my positioning.

the sermon and the markets taught me the same thing this week. you can't control the messiness of rearranging. only how you move through it.

that’s the lesson for markets. maybe it’s the lesson for everything right now.

some lingering thoughts.

• what actually helps someone without capital, without family wealth, without a head start?

• if the fed loses its independence, what happens to foreign holders of US debt? do they sell? does that push rates up, the opposite of what trump wants?

• if stablecoins win the fight in congress, what happens to community banks? do they survive? does credit dry up for small businesses and homebuyers?

• if countries keep buying gold and moving away from the dollar, where does that money go?

• what do i want to be true?

the scene: pursuit of happyness. chris gardner on the bathroom floor. holding his son. someone knocking. nothing left to control but how he holds what matters.

the read: none of us are easy to be with. sometimes the hardest person to be with is yourself.

the question i can’t shake: what do you do when everything is rearranging?

the world rearranged itself. and somehow, so am i.

i’m different. the world is different. i don’t know what comes next.

but i’m not running. i’m trusting the process.

it’s okay that everything is changing. i can handle it.

i am right where i need to be.

outcomes belong to god. reactions belong to me.

— b.

what's playing: juls, odeal, drake, premo rice.

i went ice skating for the first time this week. i’m not good at it. not even a little.

but somewhere on that rink with my arms flailing and ankles folding, i stopped trying to be good. i just surrendered to it. let myself be bad. let myself be seen being bad.

a sermon hit me later. “real faith isn’t measured by how loud you shout at the start. it’s measured by if you have faith at the finish.”

i’ve been sitting with that. the part about finishing. about staying.

and then i looked at the news and saw something shift.

the rules changed this week.

most people didn’t notice.

what is actually happening.

the us took venezuela. not through sanctions. not through diplomacy. through force. marines on the ground, maduro in custody, a new government being installed.

this is monroe doctrine energy. the hemisphere belongs to us. we’re just reminding everyone.

but here’s what most people miss: this wasn’t about oil that’s flowing. it was about oil that’s waiting.

venezuela has the largest proven oil reserves in the world. three hundred three billion barrels. seventeen percent of the global total. more than saudi arabia.

but production collapsed from over three million barrels per day in the early 2000s to under one million now. at current rates, those reserves would last eight hundred years. the gap between what’s in the ground and what’s coming out is enormous.

and that gap is the point.

most of venezuela’s oil is heavy, sour crude. dark, acidic, high in sulfur. it sits in the orinoco belt. a six hundred kilometer stretch of tar sands holding deposits that rival the world’s entire lighter oil reserves. but heavy crude is harder to refine. needs specialized infrastructure. trades at a twenty-five dollar discount to brent which is around thirty-five dollars a barrel.

the economics are brutal. venezuelan projects need eighty dollars per barrel to break even. oil is around sixty to seventy today. restoring production would require one hundred eighty-three billion dollars over more than a decade.

compare that to the competition. saudi arabia produces at three to five dollars a barrel. russia at ten to twenty. us shale at sixty-five. venezuela at eighty-plus.

at current prices, those reserves sit idle. they’re not worth extracting.

so why take them?

because you don’t take assets for what they’re worth today. you take them for what they’ll be worth when you control enough to set the terms.

and venezuela isn’t the only play. the us is eyeing greenland too. for their rare earths. they have seventeen minerals we don’t have. china controls ninety-nine percent of heavy rare earth processing. defense needs them. china’s been making moves on greenland for seven years.

but greenland is complicated. no infrastructure. no energy for mining. uranium ban. mining isn’t loved there. this isn’t a near-term solution. it’s a long game.

both plays share the same logic: position now, extract later.

and here’s what makes it heavier. the un way of doing things is in serious trouble. the us taking venezuela gives implicit permission. russia now has cover for ukraine. china is watching for taiwan.

three spheres of influence are forming. and each one is watching what the others get away with.

what’s driving this? the math doesn’t work anymore. us interest payments now exceed what we spend on defense. trump announced a fifty percent increase in defense spending. one and a half trillion. that boosts gdp short-term but adds to debt and inflation pressure. tariffs won’t offset it cleanly.

acquisition as economic strategy. not just power projection.

what i noticed.

• the us didn’t ally with a new venezuelan government. it installed one. “acquiring” not “allying.” the language matters.

• weak target first. you test permissions on someone who can’t fight back. venezuela couldn’t. now russia and china know what’s possible.

• commodities aren’t just economic anymore. they’re strategic. ai needs power and chips. brics wants a new monetary system backed by commodities. china dominates mineral processing. control of resources is leverage in the tech race, the currency wars, and defense.

• the oil isn’t scarce. the economics of extracting it profitably is what matters. reserves mean nothing if extraction costs too much.

• but that’s exactly why you take it now. when it’s uneconomical. when no one else wants it. then you wait.

how i’m seeing it.

the world order is fragile. the way of doing things is shifting before our eyes.

unilateral is becoming multipolar. rules-based is becoming control-based. and the things that matter are physical. copper, oil, rare earths, land. the stuff you can touch. the stuff that’s hard to make more of.

ray dalio has been saying this since 2023. i went back to his framework. the five forces. the three wars. he saw the capital war and the trade war coming. now we are staring at the physical war.

the market sees some of it. copper was up fifty-one percent in 2025. gold is near highs. but oil? the market shrugged at venezuela. global oversupply of four million barrels per day. the disruption barely registers.

the market is pricing physical scarcity in some places. ignoring geopolitical risk in others.

near-term, i expect more “acquisitions.” the playbook worked. medium-term, russia tests its permission to take ukraine, maybe further. long-term, china’s clock keeps ticking. their economy is struggling. taiwan is the release valve.

but here’s what dalio also taught me: things take longer than you think. he called this shift in 2023. it’s 2026 and still no war. timing humility matters.

things exist in duality. it depends how you look at it. better yet things take on what you want to see.

maybe that's what this week taught me. the ice rink and the news are the same lesson. you don't know if you're right. you stay anyway. you have faith at the finish.

some lingering thoughts.

• is this the new normal or a blip?

• if trump goes after greenland, what happens to nato?

• what if the market is right to shrug at venezuela?

• russia couldn't protect venezuela. does that weaken their sphere or make them more desperate?

• if china’s stimulus works and they buy time, does the taiwan clock reset or just pause?

• what am i willing to be wrong about? the permission structure. maybe the world pushes back. maybe the rules hold longer than i think.

the chart: copper futures (hg). new fifty-two week high, up fifty-one percent in the last year. is the scarcity already priced or is this just the beginning?

the read: china’s broken balance sheet: why i think china will invade taiwan before 2030

the question i can’t shake: what does the world look like when taking is easier than asking?

i'm unsettled. but i'm leaning in. sitting with it. watching what unfolds.

— b.