darawk

This particular person seems to be using LLMs for code review, not generation. I agree that the problem is compounded if you use an LLM (esp. the same model) on both sides. However, it seems reasonable and useful to use it as an adjunct to other forms of testing, though not necessarily a replacement for them. Though again, the degree to which it can be a replacement is a function of the level of the technology, and it is currently at the level where it can probably replace some traditional testing methods, though it's hard to know which, ex-ante.

edit: of course, maybe that means we need a meta-suite, that uses a different LLM to tell you which tests you should write yourself and which tests you can safely leave to LLM review.

There's no point in protecting one individual against an unconstitutional search that proves him guilty. The constitutional issue is the ability to have conducted the search in the first place. The only reason we suppress accurate, but unconstitutionally obtained evidence is to disincentivize the action in the future. This "good-faith exception" strikes that balance pretty ideally.

The defendants rights were violated, but there is no doubt about the legitimacy of the data, and what it implies. Police now know they cannot use this method in the future, so suppressing the evidence in this particular case does not disincentivize anything, as long as its made clear that it cannot be done in the future.

The loss of trust issue is the one I worry about most. However, it's also true that the key players who we've implemented tariffs against have had them against us for years, and we've simply absorbed that.

I don't like the mechanism he chose to implement them, or the sharpness with which they were imposed, but I do think implementing actual proper reciprocal tariffs phased in over a reasonable period of time was a good idea. And I agree with you re: the service/goods issue. Them excluding services in their trade deficit calculation is by far the dumbest part of this plan.

This has been a cornerstone of his campaigns since the beginning. I don't necessarily believe they understand the details, but I do believe they understood he would impose high tariffs, and still voted for him anyway. Tariffs weren't a throwaway line, or something. He repeated it everywhere he could.

I think it is true that most people don't understand the economic principles of tariffs, including most economists. But I do think the plan he's implemented largely comports with what he's consistently said he would do.

Yes, they are badly implemented, I agree. He is calling them reciprocal tariffs though, which implies he will lower them as they lower theirs. I don't think he's chosen a good way of estimating theirs, and so I don't think it's a particularly good policy, but reciprocal tariffs in principle are a good idea.

Even though his first pass crude approximation is stupid, it's really how other countries react, and how he reacts to them that will determine whether they behave like reciprocal tariffs or not.

There is a baked-in plan to test if they're going to work: they are formulaic, based on the trade deficit. Supposing that deficit falls, they will automaticlaly readjust downwards. I don't think the trade deficit (particularly restricted to goods, as they did it) is a good proxy for that, but it's also not completely untethered from reality.

Not retaliatory, reciprocal. Retaliatory tariffs are dumb. Reciprocal tariffs are the Nash equilibrium. Whether or not these particular tariffs are in fact reciprocal is something we could debate, though. At best they are a very crude approximation of reciprocal tariffs.

The reason for my "optimism" is that it's just as easy for him to undo this as doing it in the first place. If he keeps them in place as constructed for more than say, 3 months, without shooting them through with loopholes, then he might have a real unfixable problem on his hands, as businesses start to seriously reorient themselves. However, if over the next 3 months or so, he starts tactically peeling them back or being very "generous" with exemptions, the net economic impact could be relatively small, and maybe even moderately positive (depending on the details).

Fwiw I'd prefer the republicans win again, so my optimism is actual (not that I don't have substantial criticisms of the current admin's policies). However, it is refreshing to have a content-focused exchange on the internet about politics, so h/t to you :)

He's been articulating this policy for his entire campaign. Half the country voted for him. If you are saying that the details of its implementation are insane, then are you suggesting that Trump actually personally developed and scrutinized those details? Because I don't think any coherent theory of Trump would comport with that.

I do think a lot of the details are bad. But Trump is not exactly a details guy. And while I think they could have been better (by a lot, fairly easily), I do think they are an accurate, if crude, representation of the policy vision Trump has been consistently articulating for decades (he's been talking about tariffs and trade deficits like this since the 90s).

That's why you implement them reciprocally, to force anyone else implementing them to reduce theirs. Their problem is that the method they used to identify the tariff levels was, generously, crude. And also that it was implemented too sharply.

However, as a political tactic, the sharp implementation gives them breathing room to re-calibrate before the midterms. That comes at a real GDP cost, though.

This is a good take, but I think I can answer your question about the midterms. This is timed specifically for them. The drop we're seeing in markets right now isn't a pricing in of the tariffs, per se (at least, not yet), it's a pricing in of policy uncertainty that is going to lead to a near-term drop in investment.

However, because they are doing it so early, they will have time to recalibrate and bake in exemptions until the market / inflation is happy. Up to and including backing off of the policy entirely, if that ends up being necessary. As a political strategy, it is perfectly timed to allow Trump to "save the economy" from his own policies. This is true imo independently of what you may think about the policy as policy.

When it comes to the policy as such, recipirocal tariffs, conceptually, are designed to incentivize the overall global reduction in tariffs. So, as a headline, implementing "reciprocal tariffs" is actually favorable to free trade. However, there are some important details that they have fucked up, such as identifying tariffs with trade deficits in general, and in particular identifying them with trade deficits in goods only. That is really the component of the policy that doesn't make sense, and it is important.

Most likely, they will recalibrate and/or provide a lot of exemptions, particularly as the midterms approach. As a political tactic, I think it will work out fairly well, if they respond to the feedback appropriately - that's the big question though, and that uncertainty is the most significant reason for the market drop.

The term 'credit score' is sufficiently general to encompass literally all methods of comparing borrowers. You could certainly take issue with some specifics of how particular agencies calculate it, but the idea that there is some "alternative way of comparing borrowers" then I'd invite you to invent another formula for determining loan parameters, that does not boil down to a scalar value.

Loans involve the calculation of parameters. You can either choose those implicitly through personal knowledge, or explicitly through a scalar metric (credit score). There is no viable third option, and the first option is just a bad version of the second, in the end.

Also when we talk about preventative measures, people going to a pill mill doctors to get a refill are already addicted, but what can have a long term impact is putting in the effort to prevent people from becoming addicted in the first place, which means understanding how so many people who did not want to get addicted to opioids ended up that way.

This has changed over time. At first it was the pharmaceutical route, largely due to the shift in medical norms to prescribe opioids for chronic, not just acute, pain. Prescribing them for chronic pain is a near guaranteed recipe for addiction. However, I think things have changed in the past decade or so, with people largely moving straight to fentanyl and/or other illicit opioids. I don't think the pipeline is largely pharmaceutical in nature anymore.

Portugal

Portugal is not the ringing endorsement that it is sometimes touted as. Some indicators have improved, some have worsened substantially:

https://en.wikipedia.org/wiki/Drug_policy_of_Portugal#Observ...

Causality is hard to tease out here, but more importantly, all they're doing is decriminalizing it and offering methadone/buprenorphine maintenance treatments. And the effect on number of addicts has not been good:

https://www.washingtonpost.com/world/2023/07/07/portugal-dru...

Coca in Bolivia (I am on thin ice, I know too little, but they elected a coca grower as president)

Coca is really not anything. If you've ever chewed coca leaves, they're mildly stimulating. They're nothing like cocaine.

I think there is plenty of evidence that a considered thoughtful approach to drugs is better

Considered, thoughtful approaches are always better! The question is, what are you considering and being thoughtful about. And the fact of the matter is that the most drug-liberal cities in the US have the worst drug problems, and so do the most drug liberal countries (like Portugal).

The countries that have the fewest problems with addiction are the harshest: Singapore, China, Japan. These things are not an accident. I'm not necessarily advocating adopting policies that harsh, just pointing out that they do actually work, whereas the liberal policies fail disastrously everywhere they're implemented. I'm in favor of criminalization, but only as a tool to force people into deferral/treatment programs. I don't want to see anyone actually put in jail for using drugs, unless they fail to complete their deferral program.

That was a very long time ago. I don't think the middle east was a significant producer of Opium during that time.

Fentanyl lollipops/patches are useful in rare circumstances where someone can't swallow pills for some reason, like this one. There are other ways of achieving similar effects but its possible fentanyl is the best way to go in this situation. It is a fairly rare situation though.

I mostly agree!

I'll quibble on two points. "used as a wedge issue by foreigners", while perhaps true in some moral sense, it does not really make much sense, on closer scrutiny. To reduce things down to being some foreign imposition is to suggest that it could have been any product. But it couldn't - only opium has the special properties necessary to become this kind of product. Nobody fought a war over tobacco, or even cocaine.

It is also true that Arabia, and even the ancient Greeks ('land of the lotus eaters') were aware of and could obtain Opium. However, I'd inquire as to how it is that the primary opium growing regions of Arabia are doing lately, or say, ever.

It is true that Opium has been available to varying degrees, at various times, in various places without a total social breakdown. However, widespread, sustained, cheap availability of pure Opium without total social breakdown is, as far as I know, unheard of. The over the counter stuff in Europe and the early US were mixed with other things, as in Laudanum. Almost all of high society at the time was addicted anyway, and this was the mild form.

The Chinese discovered that they could smoke it, and changing ROA from oral to smoking is a radical step change in addictiveness. I'm not entirely sure why this didn't catch on elsewhere at the time, but the fact of the matter is it didn't, and the difference between these things is a difference in kind, not degree.

It is not a choice between continued sadism or free reign herion and cocaine dealing

I hope you're right! But I don't observe anything in the world that would support it, unfortunately. I quit because I was arrested, for instance. I want to be careful about causal meaning here, I didn't stay off because I was arrested, but it was the excipient that proximately caused and also facilitated it. It was a structural break that allowed other things to change around it.

That's not to say that the judicial system is a good way to deal with things - it's not. But the credible threat of the judicial system cannot really be done away with here without courting disaster. When dealing with highly physically addictive substances, shaping short term behavior by force is often a necessary ingredient in having any hope of shaping medium or long term behavior via therapy, life circumstance changes, or anything else.

The same thing happens with ADHD medications, the timed release dosages are supposed to last 12+ hours, but in reality they vary from 8 to 16.

Modern ADHD meds are really not comparable to powerful opioids. They are both dopaminergic, but they are night and day in terms of addictiveness. Even weak opioids vs strong opioids is night and day.

There is a large delta between the average response curve and an individual's response curve!

True! But the word "average" is actually not, itself, precise. It has at least three typical meanings: mean, median, mode. These all have quite large deltas to each other when talking about dose-response curves, and since they are curves, you would also have to pick a norm first.

There are a lot of sources of variability, but variation does not actually make it very difficult to detect improper opioid prescribing behavior. Just like the variation in people's weights would not tell you much about the strength of asphalt roads. These things are not measured in the same scale.

The majority of the pharmaceutical problem came from a very small number of people who churned out prescriptions like a literal mill. Like 5 minute appointments all day every day - not random doctors overprescribing their patients by accident. What is true is that the random doctors overprescribing provided a certain amount of cover for the corrupt doctors, for a while.

The on-ramp from minor pain or surgery to a massive, blindly-renewed, over-prescription of Oxy to an opioid addiction that spirals into street drugs

This is mostly not a thing. I have known hundreds of current and former opioid addicts. I don't think I know a single one that was "on-ramped" from Vicodin or Percocet in any truly meaningful sense. It is the case that people almost always use these first. But it is relatively rare to become an opioid addict as the result of a one-off, acute vicodin prescription, per se.

is still mediated by doctors. Until these doctors start losing their licenses for their clear and obvious breaches in their duty of care, this on-ramp will remain open.

I hope that is true! It doesn't seem like that to me, but I admit to not having carefully studied the data. Casually, there are 1.6 million opioid addictions currently in the US [1]. There were ~50,000 overdose deaths in 2019. That is, 1/1600 opioid addicts died in 2019 alone. To a first approximation, 0 people overdose annually from vicodin/percocet and other short term acute pain treatments.

It would be fairly surprising to me if (much) more than 1/1000 strong opioid users (per year) dies from an overdose. If the numbers were substantially higher than this, the epidemic would burn itself out in the population rather quickly. We can infer from this that most active opioid addicts are users of strong opioids, which are basically never prescribed for acute use. Hence, the overwhelming majority of current addicts are users of strong, non-acute opioids.

This doesn't mean there can't be some gateway effect (I do in fact think there is), but it does mean that "the problem" is mostly the presence of the strong opioids, not the Vicodin prescription for your wisdom teeth.

I'd be open to contrary data on the matter, though.

The fact that medical boards allow these doctors to retain their licenses is the core of the issue

It is an issue, and we should absolutely try to improve it. It's just unlikely to materially dent the larger issue.

I am only aware of a handful of the most obvious, blatant, and egregious pill mill operators being prosecuted. Regular doctors who simply cannot be fucked to care for their patients, and prescribe them pills so they leave their office, have yet to be held accountable.

I can personally attest to this being false. It was really quite annoying - I had to find new doctors on a number of occasions as a result, and that was ~15 years ago. Things have gotten much, much tighter on the pharmaceutical side since then. Every doctor who wrote me something was in prison or dead (from suicide, in prison) within 2-3 years of the last time I saw them, and I didn't even turn them in.

It is true that at any given time the Oxycontin prescriptions are power-law distributed, with most of the scripts being written by a small number of doctors. But this is a little bit like looking at the profits in the high frequency trading industry, or the cartels in Mexico (not to morally equate these things). Yes, there are a small number of them and they seem to make a lot of excess profits, but that does not mean you can knock them off and change anything. Nature abhors a vacuum, as they say.

1. https://www.hhs.gov/opioids/statistics/index.html

Oxycontin was, and still is, a prescription drug. I'm not sure if you're aware, but at least in the United States (where this court judgment was), it is doctors that write prescriptions. Listening to them is how the first opioid crisis happened. McKinsey dangled some shiny slides in front of Purdue, Purdue dangled them in front of doctors, and then doctors prescribed them en masse. This has, arguably, not gone well.

np,

Follow-up Q for you. What is the realistic way to prescribe opioids routinely and safely?

The simple answer is "less and with better monitoring". The first half of that was the equilibrium that Purdue intentionally shifted in the medical establishment. The cascading effects of that are what caused the modern day opioid crisis. Unfortunately, the modern-day opioid crisis as I understand it is mostly no longer related to pharmaceutical availability. So, while we should improve and lock down that supply chain route, unfortunately I don't expect it to make a large dent in the overall problem.

Are there certain formulations that have been or should be removed from the market?

Take-home fentanyl is probably unnecessary, but again, I wouldn't really expect this to be a silver bullet. The DEA/FDA has gotten much tighter on their prescribing rules for powerful opioids, but their doing so has largely coincided with the expansion of the illegal heroin, and then fentanyl markets. It is now too late to fix by choking off supply, because the market has mostly moved outside of the regulatory regime (though we should still do that, to the extent we can).

How do we reinstate the taboo on prescribing them?

These answers keep getting worse but we largely already have. We could probably do more, but if you are an MD and you are not "opioids are dangerous actually"-pilled, I think you need to go back to medical school. There was a short period in the mid 2000s where doctors were convinced otherwise by Purdue among others. Doctors who "think otherwise" today are almost without exception just outright criminals.

For EG: I got a vasectomy recently and was told to expect a day or two of pain. I was prescribed a month's supply of opioids without a single comment from the doctor on their addictive nature. My understanding is that this is how people get introduced to opioids; the pathway goes "legal scrip -> addiction -> illegal supply -> fentanyl -> death" and that's the engine of the epidemic.

Overprescription like that (which that definitely is) is bad and unfortunately common. It's hard to say exactly how much addiction is caused by that variety, though. Most serious opioid addictions that I am aware of didn't get that way from a one time moderate overprescription of things like Vicodin or Percocet. It is possible to get "mildly" addicted from a month's supply of that and when you run out you might have a slightly unpleasant day or two, but not worse than that. If the illicit market wasn't there, that 30 day supply would be the end of any binge, and that would be "mostly fine", as such. That is not an endorsement or to say that it is at all a safe thing to do, but the risk comes primarily from not wanting to quit when you run out, and having other options available.

Two things changed with the introduction of Oxycontin:

1. It started being prescribed for chronic, not acute pain. This meant that people had permanent, ongoing prescriptions for them. Which meant that people built up a very large tolerance, which led to..

2. Oxycontin is pure oxycodone, it is not formulated with an NSAID (like Percocet is). The presence of an NSAID limits the amount you can take before you get sick, and prevents you from (straightforwardly) consuming it via non-oral routes of administration, which is exponentially more addictive.

Should it be legal for the doctor to prescribe pain meds like this? (Or, should it be legal but discouraged? Is there a well-understood way to do this?) If it should be legal, should we expect the epidemic to continue? And if so, is post-bankruptcy Purdue a good thing or a bad thing?

Legal but discouraged, definitely. They are an important tool in the treatment of acute pain. They can, more rarely, be an important tool in the treatment of chronic pain (cancer / chemotherapy being a good example of a sufficiently serious condition). And finally, they are absolutely worthwhile for palliative care. For these reasons and what is now the magnitude of the illicit market, I don't think there is a lot of value in a total restriction.

Opioid addiction (I have seen a lot of it, not had it) is a social problem and is best managed with, opioids

I don't entirely disagree with you, but I have also seen enough people stop, who probably wouldn't have if that were the typical treatment, to be pretty cautious about that. There are a number of promising addiction treatments in the wings at the moment, in particular Ozempic and the general GLP-1 agonist class.

Transitional opioids like Buprenorphine are fine as a detox strategy, and maybe even fine for the medium term, but committing to them for life I think is a mistake (in most cases).

The problems stem from putting unreasonable obstacles between adicts amd their appropriate treatment. Practical legal and financial obstacles, sadistic legal obstacles, and bizarre moral obstacles

There is a lot of truth to this. It is, in fact, what I used to say when I used them. And it is and remains true. It is also true that prolonged opioid use is mostly physiologically harmless (overdoses notwithstanding). However, there are psychological elements that come with long term use that these measures do not capture, and are not fully internalized by the transaction costs (or literal costs) associated with obtaining them.

Nineteenth century society managed it with laudinum

Ask China why they fought that opium war, and how they feel about such things lately. They are still mad about it.

The Sacklers are, were, parasites profiting from that social malaise and bad things will finally happen to them. But the cause of the malaise is our irrational attitude to drugs

Agree on the Sacklers although personally I'd place more blame on the McKinsey consultants that wrote the original deck that proposed the strategy. I don't know how much the Sackler individuals personally made these decisions, but those people certainly did.

Re: irrational attitudes to drugs, I agree, but the situation is substantially more nuanced than it might superficially appear. Laudanum did used to be over the counter, as did cocaine among other things. However, these things were not criminalized for no reason - heroin wasn't criminalized in the 60s/70s anti-hippie craze, or for racist reasons in the 1930s (like marijuana).

Heroin was first criminalized for over the counter sale in 1910 - 15 years after Bayer first marketed it. Easily the fastest criminalization of a novel pharmaceutical compound in the history of the world. This is not an accident or a product of some temporary social hysteria. And unfortunately, it was also not criminalized because all of its harms were due to its being illegal.

If criminality were the problem we would expect things to get better, not worse, with the introduction of fentanyl which is far cheaper and more readily available. I could be misreading the data, but that does not seem to be working out.

Prescribing doctors are responsible for the opioid epidemic. Doctors failed in their duty of care to patients. Doctors failed to track their patient's medication usage, and failed to spot addictive behaviour. Why aren't we holding them responsible? Simply because that's hard to do?

This is a nice idea, but most Oxycontin is not prescribed by someone's doctor (it is prescribed by a doctor, but it is power-law distributed, most of it is sold by dealers). There are a small number of doctors in the country at any given time that prescribe almost all of the supply. This is not something you can readily fix with responsibility at the doctor level. It may seem like you can, because you could just prosecute "those doctors", but the problem is that the incentives are too concentrated.

That isn't to absolve these individuals of responsibility. They are responsible, and we should prosecute them legally. The problem is that we already do and always have. We should keep doing it, but I don't expect it to fix anything.

EDIT: To be clear I'm not necessarily for or against this settlement. There was a time that we might have stopped the opioid crisis at the corporate or pharmaceutical level, but that time has long since past. We could criminalize all opioids tomorrow and it would make almost no difference. Most opioid addicts use fentanyl now, and most fentanyl is produced/sold illegally. Heroin, for instance, has been Schedule I forever - the only thing that reduced its popularity was a cheaper substitute in fentanyl.

If we are going to bother prosecuting or civilly charging Purdue or its principals, it would have to be for purely punitive reasons. Corporate behavior unfortunately does not matter anymore.

Yes of course there are realistic ways to prescribe opioids routine and safely. Purdue did not invent opioids, what they did is patent a particularly strong formulation, and then work to undermine the cultural taboo in medicine around their use. That last thing was their crime.

There isn't, at this time, any other way to treat serious pain (in all cases). Opioid prescriptions will continue to be an unfortunately necessary feature of our medical system until we invent painkillers strong/good enough to replace them. I say this as someone who was personally addicted to opioids in general and Oxycontin in particular and feel very lucky to have only one dead friend as a result.

Opioids can't be banned until we can replace them with something equally generally effective. But what we should do is reinstate the taboo around prescribing them, and probably stop offering take home fentanyl prescriptions at all (it's still useful for anaesthesia, in hospitals). Perhaps more importantly, I think the cat is kind of out of the bag on prescriptions. I haven't been around that world for a long time, but it seems to me like pharmaceuticals aren't the primary entrypoint to opioid addiction anymore - people just start with fentanyl.

Dealing with fentanyl seems to me to be basically impossible. One kilogram of pure fentanyl is ~2,000,000 doses, for someone without a tolerance. This is impossible to interdict at the border. It is also fully synthetic, with no great choke point on synthesis routes. I have heard that there are other, even more dangerous, compounds like https://en.wikipedia.org/wiki/Etonitazene that are also even easier to synthesize. I don't really know if there is a good way out of this other than doing the slightly less impossible thing, which is curing addiction.

Twitter (and other platforms like it) self-impose those costs on themselves in order to maintain the feudalist structure.

No they don't. The site would not exist if they did not "self impose" those costs. By simple analogy, which self-imposed costs were paid by feudal lords?

The fact that billions of people still choose to use Twitter instead of the sensible alternative shows that the feudalist gambit (using algorithms to make people angry and turn on each other) is working.

It shows that people prefer Twitter to Mastodon. It does not show anything about why they prefer one to the other. Your reason may be a factor, but there is no prima facie evidence that it is the dispositive factor. There is quite a lot of evidence that other factors are substantially more relevant.

who cares about casualty numbers when one side initiated it by killing civilians? Israel has license to kill every single one of them until the hostage taking, rockets and murders stop. Israel didn't initiate this. Palestine did. They did so knowing this would be Israel's response. This blood is on the hands of the Palestinian leadership's intentional strategy.

Don't want your people to die? Don't organize massacres of civilians. Real easy. If your government organizes such a massacre, I'm sorry but your life is forfeit. Maybe vote differently next time.

Casualty counts are irrelevant if we aren't looking at who the casualties are and who started what.

If only there were some sort of technological method for disintermediating banks without sacrificing their advantages.

Ah well, nevertheless.

Giving a 50 million dollar donation does not mean you owe zero tax on 100 million of income. It means you owe exactly the same amount of tax on every dime you kept. Deducting charitable donations just means reducing your income by that amount. You still pay tax on all the rest.

Secondly, there are annual limits on the amount you can deduct, percentage wise. What the limit is exactly is circumstantially contingent, but you are not allowed (generally) to zero out your income even if you donate 100% of it.

That being said, I agree that donating to universities is one of the dumber forms of philanthropy.

Lol. Nobody is choking on smoke. The earth is fine. Humans may make life slightly more difficult for themselves in the next century or two due to global warming, but those difficulties will be more than offset by advances in technology, even under pessimistic assumptions.

When, for a couple of months in 1992? There are always a few people who dislike new ideas but that one was popular from the beginning because it was so easy to understand: just like ordering from a catalog but you get your stuff a week earlier and in most cases can see more info about it before buying.

I think you were in a pretty distinct milieu if you thought this was a popular idea in 1992, or any time before, say, 1998.

I worked at a web development company in the mid to late 90s and businesses were all over this concept. It saved them money having to pay people to answer the phone or open envelopes, let them do things they couldn’t afford with printed catalogs, and avoided a whole ton of problems around payment.

Yes, I imagine that the companies you interfaced with as an employee of an webdev company in the late 90s were all over this. That is not the same thing as it being generally popular.

I thought that about tablets too! I still don't really use them myself, but they just seemed like a dumb interface to me. I do understand their appeal now, though.