mattas

"Right now, none of the above analysis applies because demand exceeds supply for frontier models, and supply is limited by a lack of compute."

It gets particularly hairy because models themselves can tune their "token verbosity" to manufacture demand for compute. If compute was such a precious resource, you'd think we'd be complaining that the output was too terse.

The ability for a vendor to determine ex post facto how much a query costs is a similarly new economic phenomenon to zero marginal cost.

I'd love to be losing like Anthropic.

"Everybody's Weirded Out by AI-Except the People Who Live in Zip Codes Starting with 940, 941, 944, 950 and 951"

The fact investors think (thought) SpaceX is worth more than $2T is another example of how they aren't familiar with rocket launches.

Shareholders lost $50B within a few seconds of the scrub.

A rational person isn't willing to pay more than the present value of future cash flows. Just like how a rational person wouldn't pay $101 for a $100 bill.

Where it gets fun is that: (1) we have people selling $100 bills for more than $100 and, (2) we have people gladly buying $100 bills for more than $100.

No one forced them to IPO well above the present value of their future cash flows. Or claim their TAM was $28 trillion.

The unlock is probably somewhere in that honest, load-bearing caveat.

Thinking out loud, is productivity the ultimate macro benefit of AI? Should we expect macro AI investment to be a leading indicator of macro productivity gains?

For example, did macro investment in factory automation predict future productivity gains?

Might be more appropriate to add a "not AI" flag at this point.

Palantir has sold their flavor of "consulting" (forward-deployed whatevers) as an alternative to the old-school flavors of consulting (management consultants). Now, Alex Karp is upset because there's a new, new flavor of consulting (agents) that they now have to sell against.

Not only are your FDEs old news, but the new hotness is an order of magnitude cheaper. Tough sell.

"We dropped out of high school to build AI-powered snail teeth."

It's like a burgoo [1]. A steaming cauldron of community slop.

[1] https://en.wikipedia.org/wiki/Burgoo

I was describing this exact feeling today. I haven't quite been able to put it into words but I do get slightly physically ill. Almost similar to mild trypophobia?

Why bother learning anything when you can just use AI?

Not even close to the manhole cover speed record.

There's a disconnect between measured productivity and "anecdotal" productivity. I love this chart because it also demonstrates one of the most effective ways to increase productivity: simply reducing the workforce.

https://fred.stlouisfed.org/series/OPHNFB

Not sure how this does anything to alter the laws of physics. But I guess it's a step in the right direction.

There's another guy who I love watching (@MichaelHendriksLIVE) that exploits the heck out of biter mechanics. But it's generally in the context of difficult modded playthroughs.

"We can raise prices in two ways: (1) raise the price per token and (2) increase the number of tokens we generate on your behalf. We promise not to do (2) maliciously. Promise."

Word is just the skin is the load-bearing idea. The real unlock, if you will.

One thing that I've noticed is that there is so much content slopped out by AI that I sometimes write in that style unintentionally.

This was my reaction, as well. It's almost entirely technical analysis mumbo jumbo.

Apollo should be smart enough to know that you can't draw any conclusions from 1 month of market data (especially when there was a big, relevant IPO).

Reminds me of this.

There's a whole industry around reverse engineering tariff classifications to find ways to minimize all-in manufacturing cost.

For example, let's say you sell air purifiers.

Option 1 is to import an air purifier and pay the 25% tariff (or whatever the actual duty rate is) on air purifiers.

Option 2 is to import a widget that gets classified as a fan (with 5% duty) and import a widget that gets classified as an air filter (with 10% duty), then put them in the same box somewhere in the US.

Both are sold to consumers as an air purifier. But one of the options minimizes total cost to the manufacturer.

Outside of the "this will take your job" messaging, there is simply fatigue around just seeing it _everywhere_. The poor horse has been completely emulsified.

I can't wrap my head around how revenue > COGS but at the same time AI is being subsidized and the real cost is not affordable.

You don't price based on cost, you price based on willingness-to-pay.

So maybe labs are "overcharging" enterprises on interference (because, up til now, enterprises have seemingly had unlimited budget for tokens) and "undercharging" individuals and SMBs (because they don't have an unlimited budget).

I actually think there's an argument to be made for this to be an alternative to typical cargo ship operations.

The challenge when moving goods via ocean vessel is that everything takes _a long_ time. Loading and unloading the vessel can take days. Transit is weeks. Unloading the vessel takes days.

You have 2 options now: air freight which is crazy expensive but gets it there in a few days max or ocean freight which is relatively cheap but might take weeks. If you can cut out vessel loading/unloading you save at least a week.

"415 metric tons of goods. That means they have about five times more cargo space than an airplane, but are five times smaller in length than a typical container ship."

Not to take anything away from this (it's great), but for reference, an average vessel in Maersk's fleet can carry about 100,000 metric tons so you'd need about 250 of these to replace a single container ship.

Not sure why the article decided to compare cargo capacity of a airplane with the length of a container ship, but alas.

Ahem. "AI-powered engagement farmers..."