analyte123

A big thing that he didn't mention: Income-based repayment programs (particularly the new SAVE plan) and Public Service Loan Forgiveness (PSLF), while certainly well-intended, make the cost of college essentially irrelevant for people who can reasonably expect to benefit from those programs. And so the colleges up the bill. His 3 proposed solutions don't really deal with this. Obviously people on IDR or PSLF don't need to declare bankruptcy just because of the loans. And if the colleges had to carry the risk, the risk would be far less than imagined if these programs still exist in the same way as today: the federal government guarantees it will write off the loan of anyone who works in non-profits (which are about 10% of all jobs) for 10 years, or anyone who makes a small minimum payment for 20-25 years with SAVE.

The SAVE income-driven repayment plan is a big deal that will make the cost of college far worse. What you have to pay is 5% of income above 225% of the poverty line. A family of 3 making $80k pays $104 a month - it doesn't matter if they had $10,000, $100,000 or $1,000,000 in loans. If you are not planning to be in a very high-earning field, why wouldn't you borrow the maximum? Why wouldn't the colleges charge the maximum? This amounts to a giant subsidy.

I'm aware that many of these people couldn't pay, mostly for lamentable reasons outside of their control, but pretending nothing is wrong usually doesn't help a bad situation. Ignoring letters from the IRS doesn't make you not owe them taxes. Ignoring letters from your lender or their lawyer doesn't make you owe less money. You still owe the money and it costs you interest and time that you could use to negotiate a settlement or forbearance. Perhaps if you make your situation bad enough, it's easier to file for bankruptcy and start over, but I doubt that people are thinking that far.

In any case, it's a big stretch to call eviction "state violence". When they're even involved, the sheriff is mostly helping to enforce a private contract between the homeowner and their lender. The only conceivable alternative to this situation that still allows lending or private property to exist at all is for the police to not exist and everyone enforces their own contracts with their own violence.

During this same time period I had a project at a part-time gig which involved updating document templates and adding a few fields here and there for a legal application. The application was used by a law firm which specialized in mortgages. So I saw the details of various documents one is served in these situations: notice of default, acceleration notices, notice to vacate, and ultimately evictions and foreclosure auctions. And of course you could see thousands of these documents being generated and mailed out. For a guy like the one in the post to show up, you probably would have had to completely ignore probably a dozen of these documents for at least 6 months, not paying your mortgage all along.

Sometimes I feel like this experience, especially the idea that the entire principal balance can be accelerated if you don't make your payments regardless of the value of the house, left a bit of a scar on me that caused me to miss out on the ZIRP real estate boom II. But I am still happy with the way things are turning out overall.

70% of shelter CPI is calculated by calling random homeowners up and asking them how much they think they could rent out their house for. It was switched to this in the early 80s when the Volcker shock spiked interest rates, housing payments, and CPI. Whatever you want to say about this method’s conceptual validity, it’s clearly a lagging indicator.

Doctors don't seem to be driving the majority of the costs. After adjusting for inflation, doctors don't make much more than they did in 1984, yet the percentage of GDP spent on healthcare has almost doubled since then (and the real GDP has also doubled in that time).

For example, see [1] where a general surgeon made an average of $118,689 in 1984 and a family practitioner $84,256. This would be $358k and $254k in 2024 dollars. Today in 2024 they make on average $423k and $272k respectively [2].

[1] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4192917/ [2] https://www.whitecoatinvestor.com/how-much-do-doctors-make/

It really shouldn't be a point of pride for any kind of "researcher" to have the revelation as an adult that oil has been important in the history of the California economy. I could understand this level of naive astonishment and bitter demonization if all your education came from an elementary school textbook, but this author is supposed to be some kind of public intellectual.

From the tone of this article, you would think that oil is some kind of purely toxic, useless substance that fiendish villains decided to dig up for fun and hate, "boosterish paper[s] smirk[ing]" all along. I imagine the state of the world if the oilfields in LA and elsewhere had been "left fallow" as the author's misapplied metaphor desires. It is not good: starvation, deprivation, no modern materials or planes, extremely restricted technology development. On the other hand, those of us who still managed to be born likely wouldn't have computers and we'd be spared from having to read this.

Mineral's rover really appeared to be a solution looking for a problem. Why do you need a giant, expensive vehicle to slowly roll through your fields and see how the crops are doing when you could fly over it with a $700 drone and just look at the video? "You can't see the berries from the air!" - OK, berries are perhaps a $40B market globally. Most of the rest of the $3T+ market you can see just fine from the air or even one of the many competitors offering satellite crop imagery. The rover doesn't even do anything in the field, unlike for example the LaserWeeder implement.

If you read about agile, standups are supposed to be conducted among the dev team members who are working together on something, and primarily for their benefit, not as a status update to or for someone else. The PM or whoever is considered the "product owner" in scrum could listen in and help unblock issues, or report delays to other people, but not demand more information or change things up that have already been planned. Of course, it actually working this way is rare, although I have seen it.

I haven't really seen this, but a poorly performing product manager or product owner completely breaks the scrum or agile model (to the extent it works at all). They are assumed to basically know the domain and know what needs to be built at a high level, and the software requirements originate from them in the scrum model. If they don't know what the requirements should be or how to communicate them or how to collaborate with the engineers on the requirements, it is completely garbage-in-garbage-out. On the other hand, working with a product owner who is a domain expert and happy to help define software to solve their needs can be a joy.

It's simply not true that the only users of SQL are programmers. In fact, I have seen the disdain programmers have for SQL be a problem, where the programmers designed a Python client for data and the major users wanted SQL.

SPAs and "engineering for slow internet" usually don't belong together. The giant bundles usually guarantee slow first paint, and the incremental rendering/loading usually guarantees a lot of network chatter that randomly breaks the page when one of the requests times out. Most web applications are fundamentally online. For these, consider what inspires more confidence when you're in a train on a hotspot: an old school HTML forms page (like HN), or a page with a lot of React grey placeholders and loading spinners scattered throughout? I guess my point is that while you can take a lot of careful time and work to make an SPA work offline-first, as a pattern it tends to encourage the bloat and flakiness that makes things bad on slow internet.

A new Civic, Elantra, or Corolla all have MSRPs under $25k.

The license for this [1] prohibits use of the model and its outputs for any commercial activity, or even any "live" (whatever that means) conditions, commercial or not.

There seems to be an exclusion for using the code outputs as part of "development". But wait! It also prohibits "any internal usage by employees in the context of the company's business activities". However you interpret these clauses, this puts their claims and comparisons on completely unequal ground. They only compare to other open-weight models, not GPT-4 or Opus, but a normal company or individual can do whatever they want with the Llama weights and outputs. LangChain? "Your favourite coding and building environment"? Who cares? It seems you're not allowed to integrate this with anything else and show it to anyone, even as an art project.

[1] https://mistral.ai/licenses/MNPL-0.1.md

The author is an anesthesiologist. I don’t think we’re going to see the transplant surgeons writing with such concern any time soon.

At least in the US, 40% of wastewater solids are already used for agriculture [1]. Another 10% is used for landscaping. The same site estimates that if all US biosolids were applied to agriculture, they would supply 3.2% of applied phosphorus.

[1] https://www.biosolidsdata.org

I can't think of a better argument to have on-premises software than this article.

Are you using GPT-4 or the free version?

My go-to first screening question, regardless of where I work, is to present some JSON from a public API relevant to the domain (nothing too crazy, maybe 2 or 3 levels of nesting max), then ask the candidate to do a filter plus a sum or max - then for bonus, analysis or refactoring of their choice.

Way too often, this results in a painful slog, lots of hints from me, and, nowadays, snippets of AI generated code copied in that actually solve part of the problem but that they don't read and therefore end up breaking.

You can buy several varieties of hydroxyapatite toothpaste on Amazon now. Before 2021 or so it was a bit of a pain - a couple Japanese and Singaporean brands with spotty availability.

Edit: as alluded to by sibling comment, there are confusing claims about the exact size and structure of the hydroxyapatite ingredient used. Several brands claim that their particles (?) are smaller than those used by Novamin or by other brands and therefore more effective.

The concept of remineralizing the surface of existing teeth is totally different than what this article is talking about, though.

At a bare minimum, training and releasing a model like this builds critical skills in their engineering workforce that can't really be done any other way for now. It also requires compilation of a training dataset, which is not only another critical human skill, but also potentially a secret sauce if it turns out to give your model specific behaviors or skills.

A big one is that it shows investors, partners, and future recruits that you are both willing and capable to work on frontier technology. Hard to put a price on this, but it is important.

For the rest of us, it turns out you can use this bestiary of public models, mixing pieces of models with their own secret sauce together to create something superior than any of them [1].

[1] https://sakana.ai/evolutionary-model-merge/

The victim's parents would technically have been guilty of a felony if this happened today, by failing to lock up the firearm and the minor causing injury to another, including their own self [1]. Michigan's safe storage law was not in effect in 2022, but the truth stands that guns really need to be locked up when there are teens around. For some reason the narrative has been on 6-year-olds accidentally picking up a gun, but that type of accidental child firearm death is absolutely dwarfed by teen firearm suicides.

[1] https://www.michigan.gov/mdhhs/-/media/Project/Websites/mdhh...

The 273 btu of gas they have stored in those impressive cylinders is equivalent to 0.08 kWh. Here's hoping for more to come.

It's estimated that green ammonia costs between $800 and $1500 per ton today to produce [1]. While this is higher than conventional ammonia, it is less than how much ammonia cost in the 2022 energy crisis [2] and likely to decrease further in the future.

Massive amounts of nitrogen fertilizer are wasted because it's so cheap [3]. There's headroom for bringing back crop rotation of nitrogen-fixing crops. Nitrogen-fixing microbes are an emerging technology [4].

I am not convinced that we're all going to die.

[1] https://itif.org/publications/2023/04/17/climate-tech-to-wat... [2] https://farmdocdaily.illinois.edu/2022/09/fertilizer-prices-... [3] https://www.usda.gov/media/blog/2016/06/07/nutrient-challeng... [4] pivotbio.com

I don't know why you would take a headline as an invitation to speculate on the cause when the data itself will tell you that all of these are significantly less than the difference in drug overdoses between the US and peer countries.

I suppose the conclusion, "the biggest picture is that OER inflation is running well above pure rent inflation (see chart)...it may well the deter the Fed from cutting interest rates anytime soon" may be relevant today, but when I see the chart over time, I see that the entire concept of OER inflation is completely out of touch with reality, and that if they were observing observed rent instead they wouldn't have been so far behind the ball on inflation. Not sure if paying attention to observed rent now makes up for ignoring it earlier yet.

Most people affected by economic and social problems do not resort to theft to solve their problems. In fact, they are probably far more likely to be a victim - either directly out of sheer proximity or indirectly as stores close and opportunities fade away due to all the theft being allowed in their community.

Organized crime in Mexico would probably do it to prevent other people from using migration routes they control and reduce police efficacy. Of course, slightly rogue Mexican police or even US vigilantes also have an incentive.

The tutorials, examples, demos, and blogs for every vector DB or RAG system that I know of focus mostly on toy datasets far, far under a million tokens. So most people can be forgiven for the view that long context would kill RAG - and it really should for these trivial use cases.

I don't know if vector DB and RAG vendors don't demo their software with legitimately large datasets because they don't want to compete with their customers, or because they're not confident enough in the results, or because they don't really use their software themselves.

To give an example of a RAG dataset I have played with, it is about 10k documents and 5M tokens. Adding prior versions, expanding the coverage, or augmenting from other sources I'm sure it could blow well past 10M tokens. Maybe extremely long context models will get extremely fast and cheap, but at least for the next couple years you probably won't be stuffing this all in the context. Even for a 1M context you still have to filter, retrieve, and rank the documents. And 10,000 documents is really not a lot compared to other corpora you could imagine (e-mails, media, law, science, code, etc). But it is certainly bigger than many of the use cases being pitched for RAG - like a few hundred personal notes or a corporate wiki with 500 poorly maintained pages on it.

>The White House said threats could arise because vehicles “collect large amounts of sensitive data on their drivers and passengers"...[and] also said vehicles could “be piloted or disabled remotely” and added the investigation will also look at autonomous vehicles.

Why do vehicles have to collect large amounts of sensitive data and be remotely controllable?

That isn't a "price-cap" on electricity, it's a limit to what is essentially a tax routed through electricity bills and given to utility companies, to the tune of about $30 per person per year. The intent of the tax is to improve the grid, although its efficacy is under question, as the article mentions.

An appropriate analogy for this would be where wholesale beef markets were "deregulated", but the state government put in a program that allowed approximately half the state's population to get vouchers for a 35% discount on beef, driving up the price for everyone else (the CARE program). And the state government mandated grocers and restaurants buy certified organic regenerative beef for at least 50% of the beef they sell (the renewable requirement).