Before I would just throw prompts at the LLM and it'd end up building a pile of crap (but semi-working crap, and 100x faster than I ever could) - it was pretty depressing. Using tools like `grill-me` (or `grill-with-docs`) I feel like I'm actually building my understanding of the system and helping shape it, and the results are much better.
The fun part about that `grill-me` command is that when the questions are over, I've found that I can go right into implementation without needing to dump a PRD or some sort of broken up plan. Now this is obviously completely predicated on what you are asking it to grill you on. But for tasks that are semi complicated, it's fantastic.
LLMs are trained to convince a typical human to click the "I like this one better" on their response.
Convincing a human law professor to click the "I would prefer to deliver this response to a student" button, and to not click the "this response is pedagogically harmful" button is a different task!
I could imagine an LLM convincing a typical human to click the "I like this one better" button with flattery, or with nice-sounding platitudes, or with hand-wavey explanations that sound plausible. And in fact that's exactly what LLMs do when they go wrong - they bluff and output superficially plausible nonsense!
But these weren't typical humans, these were law professors specifically tasked with deciding which response was a better option to give to students as a canonical answer to a contract law question. So I think this is a genuinely impressive result.
Conspiracy theorists don't understand base rates and denominators. They make a list of deaths of people who seem vaguely STEM/"government secret" adjacent and forget to think about what their denominator is. Implicitly, it's "all employees (at all levels) of government labs, plus associate-director-and-up employees at STEM companies, plus all academics working in STEM fields, plus anyone who was formerly in one of these categories."
When you write it out like that, it's obvious that the denominator is ridiculously large and that it would be much more difficult to explain a lack of deaths in such a pool over a ~3 year period.
The missing worker here is typically included in the lists of "suspicious deaths" because she was an administrative assistant at Los Alamos (a government lab).
I don't know if you accidentally replied to my comment or not, but my comment was a question. I asked what's the implication? You replied here with a long text that seems to continue implying that something nefarious is going on, without specifying what the implication is (which is what I asked).
1. some people argue that the implication of this (along with other deaths of science-adjacent people) is that there's a murder campaign against scientists
2. but this "implication" is actually completely innumerate, and in reality the implication is something like "within a large population, you would expect some people to be murdered or commit suicide."
I looked at the whole context of responses to the post before my post, so (1) seemed obvious to me, and I was mostly following on with (2), but I see that it was confusing. :)
you have to have paid attention to a bunch of other news articles and their conspiracy theory dirivatives. This isn't a new item of any significant note unless you're into conspiracy theories.
You got the response you did because it sounds identical to naive-pivot conspiracy theorists who just 'ask questions' to try and expound upon the conspiracy theory.
Exactly what pool of people are we drawing from here? How big is our denominator?
- Anthony Chavez was a foreman supervising construction at Los Alamos. Retired in 2017.
- Melissa Casias (from this article) was an administrative assistant at Los Alamos.
- Steven Garcia was a "property custodian at KCNSC."
- Jason Thomas was associate directory of chemical biology at Novartis (pharma company) on cancer treatment.
- McCasland retired 13 years before he went missing.
- Nuno Loureiro was a plasma physicist at MIT.
So the denominator here has to include:
- current and former employees (at all levels, including clerical positions like Melissa Casias or construction workers like Anthony Chavez) of government research facilities
- management at pharmaceutical companies (or presumably, other STEM companies as well?).
- academics working in STEM fields (MIT plasma physicist Nuno Loureiro)
So you have to ask yourself two questions:
- how big is the pool that we're drawing these deaths from? and
- given a pool of this size, how many people in it would you expect to die in a ~3 year period?
To me the answer is really boring - this number of deaths is just utterly unsurprising. It honestly doesn't even rise to the level of being a "coincidence." It's a complete non-event.
The 2025/2026 subset is a bit more interesting than the full set. Also, only in America would 'walking off with a handgun and disappearing' be considered normal. But yes, no proof of foul play.
As someone who lives in the US I'd kind of prefer if people didn't flippantly mislead my countrymen to claim that assassinations are beneficial. I found zero sources on Wikipedia for what you claimed. Seriously, if you are too "bothered" to look up any source for an empirical claim, perhaps you should reconsider posting the claim in the first place.
Here are several articles about denial increases in the time immediately following Thompson's death:
So the "factual answer" actually is false, and you can't be bothered to even do a google search, you're just spreading misinformation for fun and spite.
The claim was not "UHC made prior authorization easier for many procedures," the claim was "rejection rates by health insurance companies plummeted." I found several sources claiming that denial rates increased following the assassination:
The evidence that car seats save lives is significantly weaker that you probably believe, as I detailed in another comment in this thread. But look: even if car seats make sense for a typical 5 year old on a typical drive in their typical car (which is a higher evidentiary burden than you might think), a mandate imposes a huge logistical tax that makes many normal things completely infeasible or impractical:
- travel with many kids (nope, physically can't carry 4 car seats plus luggage)
- using a taxi, e.g. to go see a movie (nope, can't carry a car seat into the theater)
- carpooling with other families (I'll drive them, you pick up? Nope, we'd have to shuffle car seats around.)
- rides with grandparents or other family members (sorry, we'd have to deliver the car seat to them first)
- splitting kids between two vehicles for errands (let's spend 10m wrestling car seats from one car to the other first)
The whole texture of independent childhood is altered by car seat mandates! Everything gets filtered through "is there a car seat available?". If you haven't experienced this, it's hard to describe - and I think it's absolutely a case where tradeoffs like "how will this affect quality of life?" are completely overridden because "well, if it just saves one life..."
> Car seats and booster seats significantly reduce the risk of fatal injury in crashes by 71% for infants and 54% for toddlers (1-4 years old), saving over 11,000 lives in the US since 1975
> Booster seats reduce the risk of serious injury for children aged 4-8 by 45% compared to seatbelts alone.
It's from the AI summary because it was the most quotable but the articles I found say pretty much the same thing. Seems pretty solid to me.
> If you haven't experienced this, it's hard to describe - and I think it's absolutely a case where tradeoffs like "how will this affect quality of life?" are completely overridden because "well, if it just saves one life..."
If you haven't experienced your children dying unnecessarily because it was inconvenient to make them safe it's hard to describe..
What articles did you find, exactly? What primary evidence are they basing their claims on? Many of the numbers you'll find with a google search are unclear about what they're comparing to - I believe both of the fatality numbers above (71% and 54%) are relative to completely unrestrained kids, which is not the relevant comparison.
The 45% number I specifically discuss in the other comment, but every independently reproducible study using publicly available data has found much smaller effects, around 10-25% for minor injuries and no statistically significant difference in severe injuries.
To be clear, I'm not saying "don't use car seats," I'm saying that the evidence doesn't support mandating them through age 8 (or 12!).
Our kids would be much safer if we drove everywhere at 15mph - less convenient, but it would prevent many unnecessary deaths. Unfortunately, it is impossible to do anything in the world without risk. So we're forced to balance convenience against safety every day, whether we want to admit it to ourselves or not.
It notes this, which might be pertinent to your comment regarding how the overall statistics don't show the trends you expect:
> A NHTSA study found that while most parents and caregivers believe they know how to correctly install their car seats, about half (46%) have installed their child’s car seat incorrectly.
> Children in booster seats in the back seat are 45% less likely to be injured in a crash than children *using a seat belt alone*.
That's about as much effort as I'm willing to put into this conversation. I'll finish off by saying I'm not American and these rules exist outside the US as well - I have a hard time believing so many countries would separately implement this (or similar) mandate if it was as unfounded as you claim.
We have 4 kids. Before we had our 3rd, we needed to buy a new vehicle solely because we couldn't fit 3 car seats into the back of our old car. And when traveling with kids, carrying 4 gigantic car seats plus your other luggage is not exactly as easy as you might think! It essentially rules out solo parent travel with all 4 kids. Transferring car seats between two cars, or installing car seats in a taxi, is a serious pain.
Furthermore, the evidence that car seats actually benefit safety is significantly less robust than you might think. The "mountains of evidence" that do exist for things like 70% reductions in fatalities, bizarrely enough, generally compare the rate of fatalities for car seats vs completely unrestrained kids. When you compare the rate of fatalities in car seats to kids wearing adult seat belts, the bulk of the evidence suggests essentially no difference. Fatalities happen when the forces involved are catastrophic and sadly a car seat doesn't help much for kids over 2.
Even a back of the envelope comparison makes this extremely plausible: car crash fatalities for kids 9-12 have declined by 72% from 1978-2017. If car seats and car seat laws save significant numbers of lives, you'd expect that the fatality rate for kids 0-8, who are generally in car seats, to have decreased much more. But it hasn't, it declined by 73% over the same period.
Now, car seats and boosters do seem to moderately reduce non-fatal injuries - huge spread of estimates here, most clustering around 10-25%. It's reasonable for most people to use car seats or boosters most of the time based on this alone, IMO, especially for young kids. But do they justify a mandate? IMO: no. Absolutely not.
Worth mentioning that mandates probably do succeed in one thing: they reduce the number of children born at all by at least 57x more than they prevent child fatalities. Roughly 8,000 kids per year, 145,000 kids since 1980. That's with the (unlikely, as discussed above) assumption that car seats do in fact save significant numbers of lives. But it's also entirely possible that they've prevented hundreds of thousands of kids from being born, somewhat reduced the nonfatal injury rate, and saved essentially no lives.
- https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3665046 (car seat mandates "led to a permanent reduction of approximately 8,000 births in the same year, and 145,000 fewer births since 1980, with 90% of this decline being since 2000")
Note that both the 45% and 59% estimate for injury reduction and the 28% estimate for fatality reduction all come from one research group using a proprietary data set. Everything that's independently reproducible points towards small or zero effect on fatalities and modest effects on injuries.
Yep. As I understand it, there's no statistical difference between a car seat and a seat belt over age two. We've known this for a very long time, too. But it's easy to make an emotional appeal to do something to make the world safer for children, even if what you're doing doesn't make any difference.
In the hypothetical scenario where car seats have only downsides, then of course I’m against a mandate.
There is a difference between cherry picking studies that back up your view point and how medical experts set policy though.
Experts review all of the data, and ignore outliers like a paper published in a law journal that suggests car seats are the primary reason families have shrunk from having three to two kids since the 80’s
You’ve funnily proven the point of how willing people are to put immense burdens on others in the name of safety.
There is a non-zero amount of deaths the car seat law would prevent. The burden will discourage larger families and will contribute to population decline far larger than the lives saved.
You’re not only arguing for it, you’re doing it in a way as if preventing death is such an obvious single dimension to optimize that you’re calling people irrational because they are against something that reduces fatalities.
Your same argument is what leads to prohibition and a long list of other things that suck the color out of life in the interest of “safety”.
In this conversation, you have repeatedly referred to "all of the data" and "mountains of data," yet you have posted none. Meanwhile I have posted every major study on both sides of the debate! Your argument seems to be that:
- the experts have told people to use car seats
- experts wisely base policy on "all of the data"
- therefore, "all of the data" must support the claim that car seats save lives
If we're going to discuss the question of whether experts have set policy well or poorly in a particular case, then such a strong prior on "experts always set policy well and based on the best available evidence" kind of assumes the conclusion, doesn't it?
Experts almost always set policy better than non-experts doing their own research. Especially on complex topics.
There is no point in two amateurs arguing over a topic they don't understand.
All I can do is refer to the publicly available reasoning and studies of experts, which have evidence and conclusions opposite of the amateur conclusion above.
Child car seat regulations are state laws passed by state politicians. They are not experts in any sense of the word, and generally don't bother with evidence or studies when creating said laws.
Risk tolerance is a value judgment, not an empirical fact waiting to be discovered.
Competent experts could tell you how much safer you would be if you wore a helmet to drive your car. They can't tell you how much you should value that extra bit of safety.
Personal risk is a value judgement. The government steps in when your decisions impact the life of an unconsenting third party, like a child.
Can you point to a single competent expert who recommends the average driver should wear a helmet while driving? Again, there is a difference between one study showing helmets reduce injury in crashes, and an expert reviewing the problem as a whole and making a recommendation.
We're obviously richer, too. Your grandparents had a cozy house - did they have good fresh food all winter when growing up? Could they keep food from going bad in the summer? What about indoor plumbing? These things are so ubiquitous now it's hard to even remember that they aren't just part of the basic fabric of reality.
It's easy to look back with nostalgia (and literal survivor bias - "my ancestors all survived") at the past. But if you actually look at history you will see that "what people have had for millenia" was ... pretty awful. It's an AMAZING time to be alive.
1870 is not a great span of time when OP is comparing to some idyllic village life unencumbered by urbanism. Late 19th century had many people in a “rat race” in the city, like work twelve hour days six or seven days a week in a factory type of dead-end race.
But there was something that happened later:
> For those countries with long-run data in this chart, we can see three distinct periods: From 1870 to 1913 there was a relatively slow decline; then from 1913 to 1938 the decline in hours steepened in the midst of the powerful sociopolitical, technological, and economic changes that took shape with World War I, the Great Depression, and the lead-up to World War II; and then after an uptick in hours during and just after World War II, the decline in hours continued for many countries, albeit at a slower pace and with large differences between countries.
The god knows what “sociopolitical changes” could have been about.
While there was a rat race in the city, typical farmlife was more brutal: work from early childhood and then from dawn to dusk. City dwellers had at least one day off per week, while working with animals is a job without holidays.
Say I own a spoon company. The government says "hey, I'd like to buy a million spoons from you!" I say "sure, sounds great." We sign a contract stating that I'll give them 1M spoons and they'll send me $1M.
Then the government comes to me and says "hey, actually, turns out we need 500,000 forks and 300,000 knives and only 200,000 spoons."
I say "no, we are a spoon company. Very passionate about spoons. Producing forks and knives would be an entirely different business, and our contract was for spoons."
The military now threatens to destroy my company unless I give them forks and knives instead of spoons.
You say "the voters and congress tell the military how to use utensils, not SpoonCo. Shifting the decision to SpoonCo takes power away from the citizenship."
The military can sign contracts if they wish! They can decline to sign contracts if they wish!
But private citizens can also choose whether to sign or not sign contracts with the military. Threatening to destroy their business if they don't sign contracts the military likes (or to renegotiate existing contracts in the military's favor) is a huge violation.
This article is a great example of motivated reasoning. It's comparing two wildly different numbers: the Ford Pinto number is the total number of deaths that the NHTSA found to have occurred in rear-impact crashes that resulted in a fire. The Cybertruck number is the total number of deaths from all incidents involving fire and a Cybertruck.
According to the Wikipedia article about the Pinto:
> At the time only 1% of automobile crashes would result in fire and only 4% of fatal accidents involved fire, and only 15% of fatal fire crashes are the result of rear-end collisions.
So as a back of the envelope calculation, we'd expect the total number of Pinto fire fatalities to be about 6.5x the fire fatality rate specific to rear-end collisions. Even then, I doubt that statistic would include incidents like the Las Vegas case where the man shot himself in the head while detonating an improvised explosive in his Cybertruck.
This doesn't even get into sample size - the Tesla numbers are based on only 3 incidents and 5 fatalities:
- one, a single-car accident in which 3 people died,
- two, a single-car accident in which 1 person died, and
- three, the driver shot himself in the head
If, say, the first driver hadn't had any passengers and the third driver had not been included in the sample (because it's not a collision), the Cybertruck's rate would be 60% lower. With such a small sample, it's very silly to make confident assertions about the relative risks here.
Finally, both articles are only talking about fire risks, not overall safety record. I would definitely bet that the Cybertruck has a significantly lower fatality rate per mile than a 1975 Pinto purely based on changes in vehicle safety testing and engineering since the 1970s.
Before I would just throw prompts at the LLM and it'd end up building a pile of crap (but semi-working crap, and 100x faster than I ever could) - it was pretty depressing. Using tools like `grill-me` (or `grill-with-docs`) I feel like I'm actually building my understanding of the system and helping shape it, and the results are much better.