AI investment is growing exponentially and has been for some time. Meanwhile, AI productivity is already slowing, and shows further signs that as we move forward it will slow still more. While tech companies are attempting to shovel more of it at the businesses and end-users that rely on them, there's only so much the current LLM driven systems are going to be able to accomplish without some massive change in basic operation that, thus far, doesn't appear to be happening. And while they continue to rake in investment money to continue to build out infrastructure for the fantasized giant leap forward in productivity that LLM driven systems are supposed to provide us, even *IF*, and that's a might giant if, but if they managed to accomplish *EVERY* goal the companies behind them are promising they would be able to accomplish, there is still no chance at all that they will be able to generate enough revenue to ultimate pay for the expenditure.

Do I think there is some potential in the current LLM obsession to create future productivity gains? Yes. Do I believe that anyone involved in it is holding a realistic view of how far that productivity gain will actually take us? Absolutely, positively not. This is a bubble. A bubble built on fantasy that has a foundation of bullshit mixed heavily with snake oil. And when the "correction" comes, it's going to be so much like a bubble popping that most will not be able to tell the difference.

I know, our world doesn't much care about reality these days, but reality has a funny way of reasserting itself in the most brutal and cold ways. And the longer and harder people try to fight it, the more implosively it collapses on them when it finally does come back to them. At the rate we're going, attempting to tie up the entire economy in the hype cycle, we're gonna make the economic implosion make the Titan Submersible look like a kid's game gone a tiny bit wrong.

They also get in the way of building homes.

For a while (no idea about current situation) but where I used to live in the UK you could not build a property because the Environment Agency had enacted new rules which capped water run off into watercourses, and no council anywhere in the UK had manageable plans to actually meet it with new developments, so nothing got built.

The rules were well intentioned, but came down like a hammer and stopped everything.

I want the New York Times to win and set a precedent that feeding copyrighted works into an AI without permission is a copyright violation (regardless of what the AI does with it or what output it generates). Pop the generative AI (aka "slurp up a whole bunch of copyrighted content, mix it around and spit bits of it back out") bubble completely.

Yep. CAFE-style regulation is the wrong way to attempt to reduce carbon emissions. The right way is to impose a carbon tax, then let consumers vote with their wallets and engineers work to make the right tradeoffs to meet customer demand. My guess is that consumers would choose to buy the more fuel-efficient vehicles and engineers might make the same tradeoffs... but now it would be clear that those tradeoffs are worthwhile.

Gas is pretty much exactly at its long-term, inflation-adjusted average price, and right where it was in the 1950s. Since then, it was a little higher in the 70s, a little lower in the 90s, a little higher in the early 2000s, but we're now back at the long-term normal price.

See https://afdc.energy.gov/data/1...

Whether the normal price of gas is "cheap" or "expensive" depends on your income and lifestyle, I'd think.

Ages ago I worked for a company that developed car stereos. Car companies were insanely paranoid about driver distraction. There were industry standards on minutiae like how fast song titles scroll on the screen, and a complete ban on flashing or pop-up anything.

Car companies being OK with anything flashing up on the screen that isn't absolutely critical to driving is mind boggling. All it takes is one diver glancing down for a split second to look at an ad, hitting someone walking out from between parked cars, and you have a slam-dunk lawsuit that will evaporate any money made from the advertising. Lawyers salivate at this kind of thing. Standing in front of a jury with a client all bandaged up "This callous car company thought it was more important to make money while distracting this driver by selling ads than to make sure the driver was paying attention to the road..."

This is an appropriate place for tariffs. Not ridiculous, exclusionary tariffs like we have, but tariffs carefully calibrated to offset the subsidies as precisely as possible, putting BYD's cars on a level playing field against US EVs. I have great faith in free market capitalism and dislike anything that distorts the market, but sometimes you need to use regulation to correct for external market distortions.

It's not just about which tools an AI chooses. After several months using GPT-5, I keep seeing the same pattern: "cheating" is not a binary state but a spectrum. On the bad end you have both the obvious failures, like agents selecting inappropriate or harmful tools while insisting they are doing the right thing and also something subtler and, in many ways, more damaging: the model claiming to have done research or analysis that it demonstrably did not perform.

The Stanford transparency index gives scores to various models. The models that exhibit lower transparency in one way will also tend to be less transparent in another way. In ChatGPT's case, a major structural issue is that current feedback loops are too short. The system asks the user whether the user liked the answer immediately, not whether the answer turned out to be correct or useful several steps later. But in many real tasks, I only discover the flaws in an answer a few interactions down the line. At that point there is no mechanism to assign blame to the earlier faulty reasoning. If users could give delayed, fine-grained ratings ("this part held up, this part failed"), models could learn to match surface confidence with actual reliability. It would also reduce the incentive for models to "wing it" with imaginary research because the evaluation would eventually catch up with them. And the same delay mechanism would improve safety as people are able to recognize in hindsight dangerous steps taken.

This makes me genuinely sad. The only cars left selling new are sports cars / other forms of gas guzzler.

Note to self: Check used in Mexico next time you need a vehicle.

There are three vendors Samsung, Micron and Hynix who actually produce chips.

A much larger crowd of integrators literally just scrape the bottom of the bin and glue whatever they find to a PCB. Doesn't take much to be an integrator. Some don't even have the proper equipment to properly test the memory before shipping.

How is what we're doing now progressing humanity? We're concentrating so much of build-out to make sure we keep funneling wealth, and resources, to a select few, that we're literally neglecting entire populations of people in favor of making sure we keep funneling wealth and resources to these select few. How is this progress? How is this helping anything long term? Unless the end-goal is a lowered population due to dwindling resources, I can't see how anything we're doing as a collective society is helping the average person. In fact, it seems policies across the board are specifically focused on creating situations to hurt the overall population, in favor of propping up those who already have plenty.

In short, what the fuck are we even doing?

Climate scientists didn't really overplay their hand. The media took what they reported and whipped it into panic inducing headlines. Actual climate scientists are now having to deal with a legacy of media hype on the subject on top of doing their actual jobs, because people are remembering the hype cycles, and then blaming the science rather than the shit-slingers.

Sometimes it pays to dig beneath the headlines and the media itself. I learned this lesson as a teen when they were reporting on something we were doing with our dairy cattle and showing all sorts of scare footage that bore zero resemblance to anything actually happening. Turns out, media's been hype-cycling everything they can for at least as long as any of us have been alive, and the truth behind their stories is often far different than what ends up in the reporting.

Chevy has the Malibu, at least in the 2025 year. Though, to be fair, it sounds like there's discussion it may not be made in the 2026 year.

I have no idea WTF I'm gonna do for a used vehicle in five to ten years. I don't enjoy driving giant-ass SUVs. And while I like my lighter pickups, even those are starting to disappear in favor of the king cab monstrosities that all look like penis envy relief on wheels.

I knew somebody called "vomit machine" back in high school, but I'm pretty sure the cause was he couldn't hold his liquor, yet insisted on continually drinking it anyway.