dualvariable

I asked claude to dig up the current Ford Bulletin for the engine in my truck to tell me the recommended motor oil. And it found the updated recommendations properly. I wouldn't trust google AI because I know specifically that the recommendations changed, and I don't want whatever the published specs were when the engine was first manufactured, which is out of date (and found on lots of low quality blogs). I don't even trust claude, but it gave me the URL to click on to verify and summarized it well enough that I mostly trusted that it wasn't using the cited technical bulletin and not a bunch of random AI-slop web pages.

I notice that my allergies get worse, and a theory is that systemic low-grade inflammation and histamine response may be due to diet. I don't have a lot of other conditions, but people with autoimmune conditions (e.g. rheumatoid arthritis) might see it get better/worse with dietary changes.

When I'm eating beans+rice and a bit of protein that I cook myself, my allergies are a lot lower.

When I cheat and eat "crap" I tend to start sneezing and rubbing my eyes.

There are a lot of variables, though, and cheat foods tend to be highly processed, high in gluten, with a lot of additives and made with seed oils, and hidden sugars, and are much higher glycemic index with higher glycemic loading, amongst other things. Plus seasonal allergies are just variable day-to-day which is another confounding factor. So I don't really know what the X factor is. I get similar effects on an atkins-style high fat beef, butter and cheese kind of diet, because it also cuts out most of those things.

"The bureaucracy is expanding to meet the needs of the expanding bureaucracy"

I'm not important enough for anyone in China to go out of their way to attack me. And DeepSeek has to maintain a sufficient level of trust so that users keep using their platform--they can't just act like a keylogger attacking everyone's crypto wallets or trust collapses.

If I was working on something that the Chinese government considered of strategic importance, then I would certainly be worried about it. But I don't do that.

I'm much more worried about techbros in this country using their LLMs to extensively profile me and produce something vastly more dystopian in this country than the real or imagined social credit scores in China. The people trying to convince you that the Chinese government are the people you should be worried about (as an individual in the United States) are probably the people you really need to be worried about.

This is some hopium/copium that the corporate world will trend to being naturally fair and ethical.

Amazon was always an asshole-driven company which did topgrading layoffs annually at the start of every year. It has never suffered because of that.

> The bottom half is: Black Americans Lack Assurance Police Encounters Will Go Well

My "bubble" is that I read past the headline and got more than halfway through that article that you cited.

Ask them what they think about cameras run by private companies being used to spy on them and their neighbors and see what the poll results look like.

Just because people want policing doesn't mean they want the kind of policing that we seem to be getting.

And that article you cite is a pretty good example of this.

The title is: Black Americans Want Police to Retain Local Presence

The bottom half is: Black Americans Lack Assurance Police Encounters Will Go Well

Yeah, I probably grossly oversimplified that and forgot about a bunch of other factors. Stimulus was split about 50/50 between Trump and Biden. The ARP probably was a bit too large and contributed something like 1.5% to the eventual inflation spike. A lot of the inflation spike, though, was due to supply shocks, then energy shocks from Ukraine, then "revenge spending" demand shocks as society reopened with not enough goods and a lot of excess savings. Then we had the nascent AI tech hiring boom which was pushing up costs and housing in places like SF and Seattle. By 2024 we were well on our way into the current K-shaped economy with the hypersclar buildout driving inflation due to private sector spending (even in the face of the Fed hiking rates like crazy). Really it is likely that only about ~20% or less of the inflation under Biden was due to government spending.

It is true in 2024 that there was a hell of a lot of gaslighting going on that the economy was great for everyone and that we were having a "vibecession" and praising "Bidenomics" when the K-shaped economy was already appearing, and this DNC messaging effort clearly failed. But it also didn't have a lot to do with Biden, and arguably pushing the message that implicitly accepted that Biden was entirely responsible for the economic conditions of the previous 4 years was flawed and factually inaccurate messaging. Maybe it wouldn't have been possible to argue otherwise, but I do remember a lot of Democrats who argued that Trump's economy had relatively little to do with any of his policies in his first term. I similarly think that Biden had relatively little control over the economy in his term.

Currently, though, Trump pretty well owns this economy due to what his war with Iran is doing. He doesn't own the background of what is going on with AI and hyperscalar spending and the likely bubble and the inflationary pressure and K-shaped economy (although Republican tax cuts over the past 45+ years are cumulatively responsible for the K-shaped economy and Trump owns some of that), but Trump is doing his best to just pour gas over the economy and light it all on fire.

Biden inherited COVID, which required a very accommodating policy response, but then had the AI/LLM spending/hiring boom surprise sort of overlap that, which turned out to be highly inflationary, but nobody was predicting that back in 2021. There really was no winning move to play without precise knowledge of future events. And Pandemic-era stimulus started under Trump (including Trump putting his name on the checks), and the Federal Reserve were the ones controlling monetary policy and financial sector support through the pandemic.

I don't see how you can say that he deserves "immense blame" for the inflation that we suffered. He was a passenger for most of it, and lacked a crystal ball to see the events of 2023 with the release of ChatGPT.

Unlike the current inflationary spike which is entirely due to an energy shock caused by a President starting a war of choice.

Turns out I'm already paying for that but not using it, so I guess it is time to get locked into the Proton ecosystem even more...

Doesn't matter if it is AI hallucinations or entirely human scientific fraud, the problem is the same, and the solution works fine for both cases.

If you can't validate that your bibliography is full of real articles, you shouldn't get published.

LLMs have just poured gasoline on the fire.

Yeah, I'll talk out design with AI in a brainstorming session.

Then I'll usually go and implement at least one piece of that. If I get stuck, I'll ask for some help. Then, once I'm happy with it, I'll ask the AI to review what I came up with. Then typically ask it to stamp the pattern around the codebase. And often to just iterate through writing out unit tests.

So I just did this for getting dense output from interpolants for an ODE integrator that I maintain. I did the work to make Tsit5 work by hand. I asked AI to stamp out the same pattern for DP5 and BS3, because it was just gene splicing those changes into a very similar RK integrator. I can review the diffs and see that it faithfully stamped out the same pattern with two prompts and a couple of minutes.

I'm still maintaining pretty strong contact with the codebase by doing a lot of my own programming, and fighting with the design while I'm writing that first piece of it, but then I use the AI to stamp out the mindlessly repetitive stuff.

That just seemed like the obvious way to me to go about programming with AI rather than pure-vibecoding and never touching anything other than prompts.

Also, you probably run out of tokens a lot faster if you're pure-vibecoding.

Plus you should spend some time debugging your own code. Even if AI could find and fix a bug in a minute or three that would take you 20 minutes, it is generally going to be better for you to burn that 20 minutes on trying to fix it before asking for help.

Of course, unlike another poster in this comment thread, I never cheated in college and spent a lot of time on "academic" side projects that weren't part of any course I was taking.

Once the vibecoders and cheats are done spamming a billion lines of AI generated code into industry, there's probably going to be positions for people who can (with AI assistance) sort out the mess and get production stable again.

This is one of those things that can't ever be solved with individual solutions but needs to be solved through legislation and standards, and ideally a fundamental right to privacy (and a fundamental redefinition of what privacy means when it comes to corporate surveillance of individuals).

The flip side is that the US isn't used to fighting anything other than asymmetric warfare / insurgencies / terrorism against people who can't project force and can't really hit back. We park valuable assets out in the open on military bases and installations around the world, and don't do anything to shelter them or even move them around during a conflict because of hubris.

Iran also basically just fought us to a stalemate, with an arguably long-term strategic victory going to Iran, just by being willing to absorb more punishment in the short term. Once we depleted our stocks of expensive weaponry we had to stop. We could win every fight and still lose the war.

You don't really need to be ferrying quarter million dollar horses to Europe to enjoy having horses. If you're not in an HCOL area, and not doing anything remotely fancy, you should be able to keep a horse for around $1k/month if you have a few grand buffer for vet emergencies, and you're not buying a lot of tack, equipment or supplies. You need to be into horses for their personalities, though, and willing to take on older/rescue cases, and not have fantasies of Olympic competition.

I'm particularly glad the Touch Bar got "killed by Apple".

> Wrong thing to be bothered with. This allows static analysis of every possible path the code can take. Try to do that with throw/catch. There is a reason many industry guidelines like misra, jsf, etc just outright ban hidden paths. They have been literally catastrophic. There is a reason why many modern languages like go, rust, etc, want errors to be explicitly handled.

I mean I'm explicitly supporting how rust does it over go, so I don't understand how this is any kind of rebuttal. In your last sentence you admit that rust enforces that errors are explicitly handled, and I think it does it much better than go does, and does it without littering the codebase with boilerplate if statements. The fact that you're launching into an explanation of the insufficiencies of throw/catch exception handling when that has absolutely nothing to do with what I was arguing... Well, I don't feel the need to argue with you any further. Rust also has errors-as-values and doesn't throw (at least not the common case, just like go), so the go FAQ on exceptions isn't relevant to the discussion.

And I'm just going to end the argument with that, since you seem to be a waste of time to deal with.

Yeah, this is why I'm not a Libertarian. Concentrated power in the hands of single individuals impacts the freedom and liberty of everyone else. That holds if the person is the head of government, or the CEO of a corporation.

> if err != nil is the feature, not the bug. It forces you to look at every place something can go wrong and decide what to do about it.

No it really doesn't. It litters your code with if statements that are all just about the same, except that one that needs to be different, and you go blind looking at them all and can't spot the difference. And these days people probably just type "tab" and their LLM assistant fills out the entire block incorrectly in one keypress copying the pattern from everything else.

But LLMs didn't create that problem. Having to type something never meant you had to think about it, or thousands of "sudo shutdown -r now" commands would never have been run on production databases, because typing "sudo" would have magically made someone think about it, rather than just being keyboard memory.

And the problem of reviewing the code and spotting the one error handling block that should be different from all the others is always going to be there for human reviewers.

Rust converts the common case boilerplate down into one character: ? which lets you focus on any exceptional error handling rather than a wall of if statements that almost all look alike. And the compiler can see that you're ignoring a Result from a function call and force you to explicitly do something about. Plus you can then use a monad without knowing a single thing about monoids, endofuctors or category theory, and impress your friends.

The cops have likely been doing this for decades, because the human eyeball can confuse an O for a 0 much worse than image recognition does these days.

> "It's regular practice in Colorado to list license plates with both versions, the one with 'O's and the other with Zeros in the warrant list."

That is a suspicion of what might be the problem.

And he's facing a Kafkaesque problem that in order to get him removed from the list they need to know who the warrant is for, but he also can't find out who the warrant is for. Someone can clearly figure this out and help to get it fixed, but he's been unable to talk to a person that has the ability and authorization to query the system to figure it out for him.

We really need some anti-Kafka laws in this country so that if you wind up any sort of list like this, including bans from companies like Apple/Google/Meta/etc, that you have the right to know why and to appeal, and that they must not by default assume that you're a fraudster and refuse to speak with you.

In companies, though, you often wind up with three+ massive dependency trees in your software to handle the same problem because people went and added the new hotness without deprecating the old stuff. You also find dependencies that are much heavier than necessary for the actual task at hand because the software developer was also solving the problem of needing that dependency on their resume. And then there's just the relatively tiny dependencies for fairly solved problems, like leftpad, which don't really require deps, and you can accept the maintenance burden, because not everything is an abstraction layer over chrome.

So if you just need to do something simple like fire off a compute heavy background task and then get a result when it is done, you should probably just roll your own implementation on top of the threading API in your language. That'll probably be very stable. You don't need a massive background task orchestration framework.

People might object that the frameworks will handle edge cases that you've never thought of, but I've actually found in enterprise settings that the small custom implementations--if you actually keep it small and focused--can cover more of the edge cases. And the big frameworks often engineer their own brittle edge cases due to concerns that you just don't have.

So anyway, it isn't as simple as "dependencies are bad" or "dependencies are good", but every dependency has a cost/benefit analysis that needs to go along with it. And in an Enterprise, I'd argue that if you audit the existing dependencies you will find way too many of them that should be removed or consolidated because they were done for the speed of initial delivery and greenfielding. Eventually when you accumulate way too many of those dependencies the exposure to the supply chains, the need to keep them updated, the need to track CVEs in those deps, and the need to fix code to use updated versions of those dependencies, along with not have the direct ability to bugfix them, all combine to produce an ongoing tax of either continual maintenance or tech debt that will eventually bite you hard.

I really wish they'd stop trying to suck up to me--all the "that's a really insightful question!" stuff.

I'm one of those aspy people who immediately don't trust other humans who try to fluff up my ego. Don't like it from a chatbot either.

But the fact that all the chatbots do it means that most people really crave that ego reinforcement.

Yeah, I similarly doubt that LLMs are going to directly lead to AGI just via scaling and might almost be a dead end in that direction.

But they're still quite useful tools and accelerators or force-multipliers.

And you're still going to need humans in the loop.

And I'm very worried that the capex buildout will implode once we hit diminishing returns and good-enough models can be run on substantially smaller footprints.

It all isn't going away, though, and it will still continue to improve.

Also, all the cloud models don't have to be the best frontier models, and you don't need to focus on hitting the benchmark of shrinking Opus 4.7 down to a single MBP to make significant improvements. If you get it so that an Opus 4.7 benchmark-compatible model can run in $250k of datacenter capex (and associated reduced opex for power+cooling) that'd be a massive cost improvement that makes the cloud models cheaper. And for most consumers that'll probably be good enough. You don't need to run on a $5k laptop to make a big difference.

> Designated drop off points with disabled person priority is the answer.

You have to actually put those drop off points on every street corner, though, and is the lowest priority for city planners, and gets rolled out at a glacial pace over decades--while disabled people have needs today. Having been forced to confront how well designed our infrastructure is for disabled access and put it to the test, I can tell you that a lot of existing ADA infrastructure can best be described as "performative" and a lot of it is entirely missing.

However, inference costs for entirely good enough models are likely to keep declining in the future. We're probably hitting diminishing returns on model size and training. The new generations aren't quantum leaps anymore, and newer generations of open source models like DeepSeek are likely to start getting good enough.

There's going to be a limit to how much they can raise prices, because someone can always build out a datacenter and fill it up with open source DeepSeek inference and undercut your prices by 10x while still making a very good ROI--and that's a business model right there. Right now I'm sure there's a lot of people who will protest that they couldn't do their jobs with lesser models, but as time goes on that will get less and less. Already right now the consumers who are using AI for writing presentations, cooking recipe generation and ELI5 answers for common things, aren't going to be missing much from a lesser model. That'll actually only start to get cheaper over time.

Also for business needs, as AI inference costs escalate there comes a point where businesses rediscover human intelligence again, and start hiring/training people to do more work to use lesser models--if that is more productive in the end than shelling out large amounts of cash for inference on the latest models. [Although given how much companies waste on AWS, there's a lot of tolerance for overspending in corporations...]

Over 200,000 civilians killed in the Iraq War, and we never found the WMDs.

The US and China are pretty comparable in morality, just differently.

One question the city probably can't answer is what disabled persons in the taxi are supposed to do. If you strictly enforce bike lanes the result is probably the rider needing to walk a few blocks. If the rider is disabled, that could actually be a huge burden. Since I've got an 80+ year old disabled parent with a walker this is an issue for me that does compete pretty aggressively with my support for bikes.

What is missing is a denominator and some standardization.

There could be 100,000s of people in the US who have jobs where their disappearance could be considered "concerning".

And then we need a base rate for people of similar socioeconomic status. They're probably disappearing at a far smaller rate than the general population, since they're not poor, not sex-workers, not troubled teens, etc. However, there is still a base rate, and you still need to show that it exceeds that base rate--and I kind of doubt that it actually does.

We have a large population, and over the course of a few years 10 weird things happening seems entirely normal to me.