Tim
122 posts


Dwarkesh Patel just asked the trillion-dollar question: if AGI is certain, why not bet the entire economy on it right now
“I’d be happy to buy $5 trillion of compute!”
Dario Amodei, running a $380B company, gave the answer that reveals why we’re not seeing it.
Amodei: “Bro! $300 billion of compute per year… is a lot.”
The danger isn’t being wrong about the technology. It’s being wrong about the calendar by twelve months.
Spend $5 trillion one year too early and you’re bankrupt before the miracle happens. Being right doesn’t matter if you’re dead before anyone can confirm it.
Survival isn’t optional for vindication.
Amodei: “We don’t want to go bankrupt if our forecasts are off by a year.”
Patel pushes the logical extreme. Building superintelligence? Why stop at $100 billion when the prize is infinite?
Amodei responds with reality. Deploying $300 billion over two years is already mortal risk if revenue misses projections by quarters, not years.
10x annual growth doesn’t save you when timing error at that investment scale means death. The math on scaling laws might be guaranteed. Your ability to pay bills while waiting isn’t.
Amodei: “Bro! $300 billion… is a lot, OK?”
Patel argues pure economics. If the technology works, price is irrelevant. You’re building something worth more than planetary GDP, so any investment under that is rational.
Amodei operates where bankruptcy exists. Even the fastest-growing company in AI history can’t absorb $300 billion in compute without approaching death if monetization delays by months.
AGI isn’t purely technical. It’s financial survival at scales where being correct but premature produces identical results to being wrong. You die before the technology proves you right.
Perfect forecasts mean nothing when you run out of money funding infrastructure before revenue scales to support it.
This is why trillion-dollar clusters aren’t materializing despite everyone believing AGI arrives soon. Not doubt about destination. Terror about dying one quarter before reaching it.
Building AGI means threading a lethal needle. Invest enough to stay competitive in the race. Not so much that minor revenue timing miss kills you before technology validates.
Scale conservatively and competitors outpace you. Scale aggressively and you bankrupt yourself crossing the finish line. Nobody knows precisely where that line sits, just that it’s close enough to justify massive bets but far enough those bets risk everything.
The constraint stopping $5 trillion deployments isn’t conviction. It’s the knowledge that being right means nothing if you’re insolvent when the breakthrough finally arrives.
You can win the technology race and still lose everything by spending like you’ve already won before the revenue confirms you were right to spend. And at these scales, the gap between triumph and bankruptcy is measured in months nobody can predict accurately enough to bet survival on.
English

Wait until you see Grok 5.
I think it has a shot at being true AGI.
Haven’t felt that about anything before.
Prashant@Prashant_1722
BREAKING 🚨 Grok 4 beats GPT-5 as top Vending Bench model with ~2x more sales volume and 31% higher revenue generation - outearns OpenAI GPT-5 by over $1,100 - dominates across profitability, stability, and volume - maintains sales momentum longer than most competitors Grok 4 by xAI is the undisputed market leader in performance over long time horizons.
English

@Microsoft I was cleaning my man cave and found a box of 17 year old stuff next to an old mainframe I inherited from a college I helped out. In the box was a Zune with the original charging cord still attached. I had a flood of memories that I relived just looking at this thing. 1000’s of songs that I had bought and stored on PCs and yes, this Zune. I have since lost or trashed the old PCs but I thought… what if this Zune still works. I plugged in the charging cord to a USB and 2 days later, the Zune came alive. This Zune is still Rocking in 2025. What a great piece of hardware Zune was. Nice job @Microsoft . By the way, the FM radio still works!!



English

@DailyDownfords Certainly one of the most versatile drivers but still can’t beat SVG on a road course.
English

@JackPosobiec It’s confusing. Spring forward vs. fall back. I’m for Spring forward.
English

@Commanders @SeatGeek Growing up with the Redskins all through the tough years, would have been awesome to see the old logo out there battling to go to the superbowl.
English


The penny costs over 3 cents to make and cost US taxpayers over $179 million in FY2023.
The Mint produced over 4.5 billion pennies in FY2023, around 40% of the 11.4 billion coins for circulation produced.
Penny (or 3 cents!) for your thoughts.
Sources:
jmbullion.com/investing-guid…
coinnews.net/2024/01/19/u-s…
English

@Postmates Your customer service is not what it used to be. What happened? If you pay for a meal and items are missing, your auto response now is no refund on missing items. It’s happened to me 3 times in the last 2 months. @DoorDash is much better on these issues.
English

I would like to wish our great Bitcoiners a Happy 16th Anniversary of Satoshi’s White Paper. We will end Kamala’s war on crypto, & Bitcoin will be MADE IN THE USA! VOTE TRUMP! #Bitcoin #FreeRossDayOne
English

Thanks to @HusqvarnaUSA for the HUGE donation. Brand new clothes, hoodies, rain jackets, storm jackets and more. We’re going to be heading up to the mountains tomorrow with a fuel tanker to deliver free gas and road diesel. Please bring a gas can but we do have limited supply of gas cans we are bringing as well. Location will be announced early tomorrow morning near Burnsville.
English

Super cool!
John Kraus@johnkrausphotos
Reaching an apogee of ~1400 kilometers during the @PolarisProgram Polaris Dawn mission today, Anna Menon and Sarah Gillis — both @SpaceX engineers — have traveled farther into space than any women in the history of humankind! They are amazing people. I am so proud of them.
English









