perpetualpatzer

I think it's likely the same reason people lionize Robin Hood and loathe the Sheriff of Nottingham: people see Aaron as taking from the rich to give to the poor, and Sam doing things the other way around.

I suspect for some, there's also a sense of precedent-breaking. If one conceives of Aaron's crime as the taking of the data, or rejects that LLMs sufficiently transform ideas to sever the producer's claim on its publication, one might be left wondering: "is the society who believes Aaron deserved punishment the same that thinks OAI does not?"

> high grade + long ago remains a signal

Does it? I didn't think people looked at GPAs once you're 5 or so years into a career.

Nothing, it's just hard. The problem you trip over is retention of strong employees.

Many strong employees are ambitious and need the prospect of progression to be happy with their job. When growth slows below the employee ambition rate, not everyone can be sustainably paid more. Then you're faced with a choice of removing people to make space or employees being unhappy.

You can let this equillibrate naturally. The employees who are both ambitious and skilled enough to land other jobs will disproportionately be first out the door, leaving room for others. But this reduces the average talent level, and hurts company prospects, so many companies choose other paths. Some proactively fire the bottom 15% every year. Some heavily differentiate pay to make up for lack of promotions. Some constantly reorg and fail to find spots for employees in internal interviews. Offering a paid out for those who were thinking about it anyway seems a uniquely humane solution.

It would be interesting to see this argument stratified by some measure of college preparedness like test scores, or admission to a 4yr school.

If -- as the article seems to imply -- most of the effect is explained by what schools and programs poorer students are pathed into, it seems plausible that the failing is at the point of pathing (primary/secondary education).

So does English. Well, sorta.

Neither of those answers seem at all dishonest. They just skip past the diagnosed cause to focus on the impact.

Maybe the boss being a raging ass is the root cause of the team moving in the wrong direction. But that may be debatable, and knowing that doesn't answer the interviewer's question. The more relevant piece is that you and the director disagree and that you don't want to waste your time working on something you aren't inspired by.

Why are you leaving your current job questions are typically inartful checks whether there's "something wrong with you." Refusing to meaningfully answer can come across as, "I'd prefer not to discuss whether there's something wrong with me." Not explicitly damaging, but certainly not confidence building.

The most useful answers I've heard in the wild were dispassionate, one sentence expressions of why <objective fact> made the existing job irreconcilable with your <valid need>. Done well, that shows the hiring manager you're able to approach conflict constructively, and gives reason to believe the bad fit is unlikely to recur in this role.

AskAManager[0] also suggests "I’m not actively looking, but I saw this job and was really interested because of X," as a reasonably broad spectrum solution to this problem. (Though after just a year, I as a hiring manager might still worry you are flighty).

[0] https://www.askamanager.org/2023/02/how-do-i-tell-interviewe...

Not OP, but also used the API for music discovery, crawling the related artists network from playlists I'd created to find highly connected nodes. I found that strategy was very efficient at filling blind spots in major genres I listened to, and reasonably good at exploring random niches, but generally wasn't the place I found songs that felt truly fresh to me. It felt like it got noticeably worse about 5 years ago, when I perceived them to have ramped up the payola, but that might just be in my head.

The major costs you're missing are marketing and "plate" (up front cost to produce the content). Those make up most of the total costs. For textbooks, the decision makers are professors (so door-to-door sales to get their attention), and the market is pretty small, especially for upper level content (so few units to amortize fixed costs over). Print, paper, and binding are cheap, say ~$10-12 average for a textbook. Typically, distribution channel takes a 20-25%, depending on channel partner, and many colleges mandate that sales go through the school bookstore because they get a cut, so publisher's website isn't necessarily a viable option (without a lot of student marketing). Author royalties run ~13-15% of revenue, and editors hit plate expense (they're publisher employees, so not a variable cost like authors). Textbook publisher Ebitda margins wind up running 20-25%, but most publisher's pay a lot of interest expense, partly because the major costs are up front, and partly because there's been a lot of PE ownership. Net margins can be tight as anyone else's.

Source: worked for a plaintiff publisher in this case. Think Pearson, Cengage, and MHE all publish financials also.

The chart you linked stops before it gets to the cohort discussed in the article. The youngest cohort shown is the 25-34 cohort in 2016 would have been born in 1982-1991 ("gen y"/"millenials"). The economist article is discussing the cohort born starting in 1997. It seems to think the chart gets rosier for the young.

When a company goes into bankruptcy, they still have assets (inventory, a brand, real estate, customer contracts) -- those things just belong to their creditors now. The new owners need _someone_ to manage selling all those things to recoup some money.

More often, though, the business is still worth more than the sum of its parts. It just wasn't making enough to pay the loans it'd taken out. Since the loans get eliminated in bankruptcy, the new owners might decide to just keep running the company (and want a new CEO, who they think will do a better job.

This chart looks like it's mostly geopolitics. The Brent-Ural spread opened pretty significantly in Jan.

[0]https://www.neste.com/investors/market-data/crude-oil-prices

It does, but it can be expensive enough as to be prohibitive [0]. For example, offices tend to have fewer water hookups, fewer breakers (meaning high construction costs to repurpose) and wide floors (meaning lots of windowless units).

[0]https://cre.moodysanalytics.com/insights/cre-trends/office-t...

They can be, and that's a statement about the distribution and consistency of skill in those activities. I see little reason to believe those distributions apply to SAT taking or general capacity to learn (which is what you're really aspiring to measure).

If we set up an ELO league for best-of-three rock paper scissors, I'd expect score to diverge over sufficiently many hours, but would give almost even odds to a 99th v 93rd percentile candidate.

Stay bonuses [1] are probably the most common case. It's actually pretty common in a merger, especially one with a long period between announcement and integration, to hand out parachutes of various metals to key employees.

If you're the CEO closing a deal where the buying party is overpaying by $15B, they really really don't want you to quit in the middle.

If you're a rock star category manager, they still want you to stick around because you might get the job after in the combined entity, and even if you won't, who will they be able to hire to manage the category while they're doing the deal?

[1] https://www.forbes.com/sites/lizryan/2015/03/28/what-is-a-st...

Not much. Most peers and hiring managers stopped paying attention to these ratings when they got into college, so the market's perception of your degree is some blended average of the programs' ratings over the past ~40 years. By the time rankings after you matriculated represent a meaningful portion of the average, your alma mater is no longer a particularly relevant part of your resume.

>I'm sure that the answer is "both of course"

IANAL, but my understanding is that the default rule in the US is "each party pays for their own lawyers, regardless of who wins"[0]. Of course, the specifics would depend on the actual terms of the settlement agreement.

[0] https://en.wikipedia.org/wiki/American_rule_(attorney%27s_fe...

The discovery and cost structure problems are points a lot of people miss. Textbook publishers are selling professor time in the forms of content discovery, basic homework creation, and basic lesson planning.

One expansion on the principle agent problem: college bookstores handle. Many bookstores rent their land from the college with the agreed rent of $X +Y% of gross. This creates a disincentive for the institution to spend time negotiating with publishers (or ensuring faculty are getting enough use for the materials they assign).

True, but average citizens don't _choose_ to use heavy trucks. I'm not sure of the relative costs of road repair v. fuel, but if you funded roads entirely through a tax on the fourth power of vehicle weight (if that's the right proxy for repair costs incurred), Walmart might find they could reduce total costs in the system by using fewer large trucks and shift their logistics mix.

Bayesian Methods For Hackers is a very popular one.

[0] https://github.com/CamDavidsonPilon/Probabilistic-Programmin...

Arguably, it's good for both. Buyers have more quality inventory to choose from and can purchase a home with lower risk of getting trapped in it permanently. Sellers get faster sales with a higher floor on prices.

I think it's less taking a loan to buy golf clubs and more "feeling less financially stretched makes people more willing to spend money."

Example: Joe just had a kid and was going to buy a house in a good school district no matter the cost. With a higher mortgage rates, he'd have wound up house poor for a few years. With today's rates, he has a comfortable savings rate. Since he's not scrimping, he decides this month he'll buy that putter he'd been eying.

The linked pdf contains notes explaining HFT circa ~2011. The About section explains it's sourced from the author's blog, and was also used as lecture notes for an undergrad course. I've not read the whole thing, but at a minimum, the brainteasers contain some good nerd-sniping content.

For those wanting more recent additions, the author seems to have more updated thoughts here[1], as linked on the LinkedIn profile to which their former blog redirects.

[1] https://blog.headlandstech.com/2017/08/03/quantitative-tradi...

This article outlines 2021 timing, which is typical, from my understanding: https://www.immi-usa.com/h1b-lottery-predictions/

Your transient point is very appealing to me. The line of thinking that people move more-> lower incentive to build custom-> cheaper finishes/less creativity feels logical and can be tidily appended to the Enrico Moretti argument that economic opportunity has concentrated in the cities-> more people detach from their hometowns to chase opportunity.

There are two pieces of information that I struggle to integrate into this model and would be interested in how others reconcile:

1) US census data suggests (surprisingly!) that the % of residents moving each year has declined significantly since 1990, and before that was about the same back to the first data collected in the mid 40s (though with some collection gaps) [1]

2) Anecdata: most of the new builds I see around me (Boston MSA) are luxury apartment buildings or way out in the suburbs. The former aren't quaint, but are decidedly custom architecture. The latter, I stereotype as "folks settling down in the suburbs". Maybe McMansions, but I wouldn't expect transient. This one might just be a sampling error on my part.

[1]https://www.census.gov/newsroom/press-releases/2016/cb16-189...

I imagine the answer here varies a lot by elo of the training corpus because skill effects the probability input. In the 500s, 3. Qh5 to scholar's mate is both potent and very probable. In the 1300s it's just a mistake. It would be really interesting to see the overlap between high and low elo traps.

No, payments to authors are percentage-based, so they deflate down with the price.

It's just a form of price discrimination. If you're paying and not urgent, wait a couple weeks and the price is $10. If you've gotta have it now, or don't care about the money (e.g. work will reimburse), the price is $100.

They would be 90% smaller (though hopefully on higher volume). From the S-1:

>We incur content costs in the form of payments to our instructors, generally determined as a percentage of total revenue generated from their content.

This is a pretty typical structure when unit costs are low. It gives the publisher the ability to price as they please to optimize revenue, and because the inventives are aligned, authors don't need to know customer willingness to pay.

Can you elaborate more on what you mean by "define delivery"?

> You can tell a son that all their male ancestors had a son

This seems off... by counterexample, my grandfather had no sons, but three daughters (one being my mother).