HNHacker News
TopNewBestAskShowJobs

sxg

1,702 karma · joined April 29, 2016

undersample.substack.com github.com/sxg
submissionscomments
sxg··on iTerm2 and AI Hype Overload
This isn’t the use case for the majority of people. Most are amateur developers who copy/paste the first thing they see on StackOverflow without understanding what they’re doing or reading the explanation because they just want to move forward on whatever project they’re actually working on (ex: building a REST API but they can’t install the necessary Python packages the YouTube tutorial they’re following uses). LLM-powered guidance here is tremendously useful, and if the person is curious, they can ask the LLM to explain the command, which it’s actually very good at based on my experience with GPT-4 and GPT-4o.

I think saying there’s “no confidence in the answer provided by an AI” is an overstatement and underestimates the value AI can have for the majority of users because that statement overestimates the value of “a reputable source that provides not only the command I’m looking for, but an explanation of the syntax” for the majority of users. Reputable sources and thorough explanations are great in theory, but very few have the patience to go through all that for a single CLI command when they want to make visible progress on the bigger picture project they actually care about.

sxg··on Hot take on OpenAI’s new GPT-4o
There’s no analysis and we have incomplete information. This is just a reactive take on releasing a faster model. We don’t know if a higher quality model exists, we don’t know the nature of OpenAI’s connection with Apple and iOS, and he’s dismissing improved model speed with reduced cost without explanation.

The post is labeled as a hot take, so I’m not bothered by it, but it also isn’t worth reading. Glad it was short.

sxg··on The business of wallets
There is an alternative: use cash.

> Most transactions are repeat business and require no trust intermediary.

I'm not an expert in this field, but I know enough to be humble when it comes to cybersecurity. I don't think fraud prevention is that simple.

sxg··on The business of wallets
This argument gets brought up repeatedly on HN (most recently on Patrick’s last post about credit card rewards). The counter argument is that there are costs associated with securing transactions and fraud prevention that do benefit all credit card users. The “tax” does result in a downstream benefit.
sxg··on Ontario family doctor says new AI notetaking saved her job
HIPAA trumps the warrant. HIPAA is serious when it says patient data can only be shared for treatment, payment, and hospital operations purposes. Law enforcement is not an allowable reason to disclose patient information without their permission.
sxg··on Ontario family doctor says new AI notetaking saved her job
Yes, I do believe what you're telling me—the state of healthcare tech is definitely leagues behind general consumer tech. However, I do think this is a meaningfully different class of issue when it comes to patients' perceptions and actual harms. Patients are afraid of having their health data used against them. For example, revealing medical conditions to potential employers, revealing health information to friends/family, etc. There's a growing mistrust of healthcare institutions in recent years, and there are unfounded accusations of healthcare institutions selling data for financial gain like social media companies and even the DMV (https://www.caranddriver.com/features/a32035408/dmv-selling-...). This class of nefarious patient data privacy/security negligence effectively doesn't happen. I've treated patients who are illegal US immigrants, I see patients who use and possess illegal drugs while in the hospital, but they're not reported to anyone. They're simply treated and discharged. Unfortunately, a growing number of patients don't believe this is the case, and we see substantial disparities in levels of care provided to these patients who fear healthcare.

I'm not at all dismissing how terrible it is that healthcare tech companies can be lax with patient data. This absolutely needs to be better! But at the same time, this sounds more like incompetence than active malice. Practically speaking, a patient is extremely unlikely to experience actual harm because a developer accidentally took patient data home on a personal laptop. Although, I would love to hear more about what kinds of violations you've seen in your time in health tech? I work with third party vendors from a healthcare institution, and I absolutely want to figure out how to fix this.

sxg··on Ontario family doctor says new AI notetaking saved her job
No problem—glad to participate! There's a lot of cynicism that leads to misinformation about how healthcare works, so I'd like to clean that up. Let's attack and fix the broken parts of the system, but we should praise the working parts. I think patient privacy/security is one of the few things the US gets mostly right about healthcare.

Regarding the privacy policies: these are created by the legal department and physicians in the department are told to distribute them and get signatures when necessary in order to do things by the book. However, your rights are inalienable and protected regardless of whether you actually receive the policy and sign the appropriate box. If you don't receive the policy, the healthcare institution is on the hook and could face a fine if reported to the DHHS. Things could absolutely be done more efficiently and clearer for patients, but there's a fear in changing things ("if it ain't (horribly) broke, don't fix it"). Trying to improve how privacy policies are disseminated and patients informed could result in an inadvertent violation of HIPAA that results in large fines. So healthcare institutions are disincentivized from trying to improve things here.

I reviewed the patient privacy policy for a few large institutions in the US, and it all seems to support what I'm saying. For example, here's NYU's policy on business associates: https://nyulangone.org/files/business-associates.pdf

NYU has additional policies here: https://nyulangone.org/policies-disclaimers/hipaa-patient-pr.... UCLA Health has similar policies here: https://www.uclahealth.org/privacy-practices. Every institution has essentially the same policies as they're all just a reflection of HIPAA.

The only ways in which patient data can be shared with others are if (1) they're involved in your treatment (e.g., your doctor at another hospital), (2) payment purposes (e.g., insurance), (3) health care operations (e.g., third party vendor software like EMRs, PACS, etc.) All are required to be HIPAA compliant if they're covered entities (i.e., healthcare institutions) or sign a BAA with a covered entity that essentially puts the same HIPAA requirements on them. A violation again results in massive fines, C-suite level firings, and expensive legal fallout.

sxg··on Ontario family doctor says new AI notetaking saved her job
My email is in my profile page. And if you have truly found that the institution is sharing protected health information (e.g., even just names and date of birth) with third parties who have not signed BAAs, that is a lawsuit worth tens of millions plus government fines of $50,000 per piece of compromised data per patient. I highly suspect that there's some misunderstanding or miscommunication here.
sxg··on Ontario family doctor says new AI notetaking saved her job
It's mind boggling because I highly doubt it's actually true. I'm not sure where the OP is getting that info. Patients can't waive away HIPAA privacy/security rights.

I think the OP is assuming that when healthcare institutions partner with third parties, those third parties are not required to uphold HIPAA. If that's his/her belief, it's 100% false. Third parties associating with healthcare institutions have to sign business associate agreements (BAAs) that require them to uphold the same standard of privacy/security regarding patient data as the first party healthcare institution. There are severe financial penalties for violating HIPAA, and every healthcare institution I've been a part of takes this extremely seriously.

sxg··on Ontario family doctor says new AI notetaking saved her job
Wait—this doesn't make any sense. I'm a physician and have a lot of experience dealing with protected health information. Third parties are required to sign a HIPAA BAA and obligated to uphold privacy/security standards equal to that of your physician and hospital. Can you provide some specific examples of the third parties you're talking about?

MyChart itself is a component of Epic (the EMR) and is absolutely HIPAA compliant. Every healthcare institution I've worked with has taken HIPAA and privacy/security regarding patient data extremely seriously. Non-HIPAA compliant vendors are an immediate non-starter and don't even enter discussions when looking at new products.

sxg··on How to Use JSON Path
This is exactly the type of thing that LLMs are very good at generating and explaining for you. I've done this countless times to create and understand regex patterns.
sxg··on Whistleblower Josh Dean of Boeing supplier Spirit AeroSystems has died
I agree with you in that the population size is the key question here. However, I have two issues:

First, otherwise healthy people don't just die from stress. Stress can sometimes exacerbate underlying health issues and lead to a long, downward spiral in health that can result in death, but it does not happen in a matter of just weeks or a couple months. It also does not happen in people without underlying health issues.

Second, while a mortality rate of 0.001/yr is reasonable for middle-aged men, that assumes we know nothing about them or their deaths—that isn't the case here. John Barnett's death was a suicide. According to the CDC, there were 14,668 suicides in the 45–64 age group in 2021. The 2020 census shows that there are 85 million people in the US in that same age group. The suicide mortality rate comes out to 0.00017, which is about an order of magnitude lower than your estimate. Josh Dean was otherwise healthy from what's being reported. Given his age and state of health, his 1-year mortality rate is also likely substantially lower than your estimate

sxg··on Whistleblower Josh Dean of Boeing supplier Spirit AeroSystems has died
I'm baffled by this comment. It's exceedingly rare for two middle-aged and otherwise healthy people to die without warning.

> With so many people in a position to notice and speak out about Boeing's issues

So many? Both were whistleblowers of which there are allegedly no more than 32 in the last few years.

sxg··on Whistleblower Josh Dean of Boeing supplier Spirit AeroSystems has died
> Hospitals are supposed to prevent this from happening, but it is hard.

Physician here. Hospitals do work to prevent the spread, and it is hard—so much so that it's nearly inevitable. MRSA is a naturally occurring part of the respiratory tract and skin. How do you stop that? Also important to note that those who do acquire and die from MRSA infections tend to be relatively ill to begin with. Usually those patients are admitted to the hospital for some other serious issue and then develop a MRSA infection either from their own body or the hospital. While you can mitigate the latter, you can't do anything about the former.

sxg··on Windows 10 reaches 70% market share as Windows 11 keeps declining
Windows is good enough for the vast majority of people. Businesses simply care about achieving their broader goals (i.e., selling their own products or services) and want to minimize effort on tools required to do that. The vast majority of people know how to use Windows and Office, and although they both suck in various aspects, they're familiar and good enough.
sxg··on How ECMO is redefining death
That's definitely not the main problem. Risk of intracranial hemorrhage/strokes and other complications make it problematic.
sxg··on This is a teenager
> This should be the measure of our country

The most important thing about a metric is that we all agree on how to interpret it. While money may not measure exactly the right thing, we can all agree that $50 is $50 regardless of our age, race, gender, upbringing, etc. Take a look at the other comments on this thread. There's hardly any consensus on what this visualization shows. If this was how we measured our country, we'd get nowhere because there's no agreement on what this data means or what to do about it.

sxg··on Developers, what marketing strategies work on you?
Definitely not a fan of the Databricks site, but the curl site also isn't very informative for a brand new user.
sxg··on Anatomy of a credit card rewards program
Agreed. Lots of parentheticals that distract from the main point.
sxg··on Ask HN: Do you also marvel at the complexity of everyday objects?
Yes! You'd probably enjoy this blog: https://ciechanow.ski
sxg··on Boeing Whistleblower: "If Anything Happens to Me, It's Not Suicide"
We don't know that all the laundry has been let out. It's also possible that there could be criminal charges against some of Boeing's leadership if this guy were to testify against them or if he had more info.
sxg··on FDA clears first over-the-counter continuous glucose monitor
> The parent comment mentioned feeling tired after consuming something with a lot of sugar, was that really a mystery prior to using CGM? I doubt it.

They may not have known just how sugary that item was since most people don't bother to calculate the sugar content of everything they consume, but the CGM puts a number on it. That experience could be enough to change the OPs behavior. I agree that most people may not care to make behavioral changes in response to more data, but there are definitely other people who will.

sxg··on Ask HN: What alternatives exist to a Mac mini but for Windows?
I have a work-issued Lenovo ThinkCentre M Tiny...it's so infuriatingly loud! The fans seem to spin up at completely random times that have nothing to do with workload, and they're incredibly loud.
sxg··on Brexit has failed for UK, say clear majority of Britons – poll
Not at all the same situation. Although imperfect, the UK has an actual functioning democracy while Palestine is an impoverished, semi-autonomous, largely refugee state partially controlled by Israel.

This video explains how the Palestinians voted for Hamas in theory but not in practice: https://www.youtube.com/watch?v=pJ9PKQbkJv8

sxg··on Ferret: A Multimodal Large Language Model
By "A series" are you referring to the Nvidia A100 or the Apple A-series iPhone/iPad chips? If the latter, I don't think you can draw that conclusion. Training has memory and processor requirements that are very different from inference. You don't need iPhones and iPads to train models—you need them to run models. These are two very different things.
sxg··on Spotify's push into audiobooks sparks concern among authors
We're already there. I use Readwise Reader and often take advantage of the text-to-speech feature for listening to articles from the Economist. The Economist already provides an audio narration of the article by a human, but some of the narrators' British accents are a bit too thick and difficult for me to understand at 1.5x. Readwise Reader's text-to-speech does a much better job at 1.5x for my American English brain. AI-narration isn't perfect—here are some issues I have encountered:

- Section headers are read just like regular body text, which is confusing

- Not great with reading numbers. Narrating a year is very different from narrating a monetary value.

- Parenthetical statements are read just like regular body text

sxg··on Positive labor sorting has been increasing for men in the last several decades
Sure, I guess I’m conflating society with government. Yeah, doctors and lawyers make more money than other professions, but those incomes are a result of innumerable positive and negative incentives (eg, cost of education, opportunity cost of education, work/life balance, “meaningfulness”). It seems nearly impossible to intentionally modify those incentives in a specific direction without causing unintended consequences elsewhere.

Example: society wants more quality teachers, so society chooses to increase the federal education budget to increase teacher salaries. Consequently, more highly educated people choose to become teachers, however, this unintentionally exacerbates the doctor shortage and further increases corporate lawyer salaries due to decreased supply of lawyers.

sxg··on Terry Gross and the Art of Opening Up (2015)
Are there any interviewers that don’t violate these pet peeves?
sxg··on Positive labor sorting has been increasing for men in the last several decades
How is it that society is preferentially allocating resources to creating lawyers rather than engineers or researchers? It seems to me that people chose to become lawyers, engineers, or researchers of their own free will. Sure, many researchers are government funded, so you could draw more people to that field with more resources, but I'm not seeing how society is pushing people into law over engineering at least.
sxg··on Volkswagen will bring back physical buttons
A lot of survey data supports this sentiment, and I personally agree as well—I wouldn't ever buy a non-CarPlay car in 2024. However, car manufacturers that remove CarPlay may be able to decrease prices, and their target customer is very price sensitive. I think a lot of diehard CarPlay users would consider buying a non-CarPlay car if it were 5–10% cheaper overall.
← PreviousPage 3 of 11Next →