There are many use cases for which the price can go even higher. I look at recent interactions with people that were working at an interview mill: Multiple people in a boiler room interviewing for companies all day long, with a computer set up so that our audio was being piped to o1. They had a reasonable prompt to remove many chatbot-ism, and make it provide answers that seem people-like: We were 100% interviewing the o1 model. The operator said basically nothing, in both technical and behavioral interviews.
A company making money off of this kind of scheme would be happy to pay $200 a seat for an unlimited license. And I would not be surprised if there were many other very profitable use cases that make $200 per month seem like a bargain.
So, wait a minute, when interviewing candidates, you're making them invest their valuable time talking to an AI interviewer, and not even disclosing to them that they aren't even talking to a real human? That seems highly unethical to me, yet not even slightly surprising. My question is, what variables are being optimized for here? It's certainly not about efficiently
matching people with jobs, it seems to be more about increasing the number of interviews, which I'm sure benefits the people who get rewarded for the number of interviews, but seems like entirely the wrong metric.
Scams and other antisocial use cases are basically the only ones for which the damn things are actually the kind of productivity rocket-fuel people want them to be, so far.
We better hope that changes sharply, or these things will be a net-negative development.
Right? To me it's eerily similar to how cryptocurrency was sold as a general replacement for all money uses, but turned out to be mainly useful for societally negative things like scams and money laundering.
It sounds like a setup where applicants hire some third-party company to perhaps "represent the client" in the interview and that company hired a bunch of people to be the interviewee on their clients behalf. Presumably also neither the company nor the applicant disclose this arrangement to the hiring manager.
If any company wants me to be interviewed by AI to represent the client, I'll consider it ethical to let an AI represent me. Then AIs can interview AIs, maybe that'll get me the job. I have strong flashbacks to the movie "Surrogates" for some reason.
But they absolutely can. Go see Friedemann Freese Copycat, which has zero unique rules: Everything is stolen from a popular game insiders know. Hell, it even has a misprint on purpose, matching a famous misprint! You can teach it in 5 minutes to the right people, just by reference.
The modern, kickstarter heavy 3 hour monstrosity just can't assume that the buyer has played all the games that have the same mechanics they are basically lifting from elsewhere, while only explaining the 2 or 3 places where they are doing anything interesting. But when you go through lives rules explanations among people in industry, half of the rules are really handled by reference, because you know what is going on. With some designers, the rules are almost unnecessary, as the played aids and the graphic design do 90% of the work. I've played that game with a certain designer: He sat there to answer questions, but he didn't even hand us a rulebook, or provide an explanation. Just the components in front of us and 'figure it out', as an experiment on the game's learnability
Kickstarter games tend to have far more design than development, as the success of the kickstarter campaign has little to do with the game being any good, but with the material available at campaign time being attractive. So overproduced games with way too much art, and way to many systems, end up doing far better than, say, a simpler game that would have gone through a more traditional euro publisher 10 years ago.
If you talk to people in industry, they realize that this is making the games worse, but it's making them sell better. Just like today, just like with a movie, high sales don't come from great mechanics, but franchises or licensing. Your superhero coop game is going to do much worse if it doesn't have a license. Why do we have so many boardgames with video game licenses? Because it sells. There's way too many games in the market right now for even industry insiders to good track of most offerings. Buyers working for large internet game stores are overwhelmed.
So no, game designers are 100% understanding their task, which is to make a game that gets backed and is profitable. Cutting half the game after you told backers what you were doing is a disaster, and it's all those extra rules and extra boards that sold the game in the first place. The fact that the game is shelved after 3 plays is not a problem for them.
How many times do Kickstarter backers have to get burned before they understand these ideas and switch to backing things that are actually likely to have good gameplay?
The issue with software documentation is very different than a boardgame: The boardgame is static, changes to the boardgame are ultimately changes in the documentation, and people will not pay for a game they can't understand, while people end up using systems that are undocumented.
There's good documentation in the software world, but it's always for systems shaped as to have all the incentives align to having good, current documentation. Stripe has good docs, because that's part of being able to onboard people: Bad docs cost money. Postgres has good docs because it doesn't change that much, so the good documentation stays useful, and they have a quick loop between finding errors and fixing them.
Your typical internal project has awful documentation because it's nobody's real job, the requirements change constantly, and not enough people would use the documentation in the best of times to make good documents a worthwhile investment.
And if you ask me, the SPI wargames aren't exactly pinnacles of good rulebook writing. If anything, the similar shape came from the games themselves having similar bones, more than because the system was good, or because standardization helped.
In computer science, we can squint and claim that a functor is just something that has a .map function, taking a function. Option, Promise, List... all functors.
The fact that functor is a concept, and follows specific laws, allows us to have other abstractions that work across all functors without having to repeat ourselves over and over again. See, for instance, Scala's for comprehension. it offers different iterating capabilities over basically anything, and it will give you more capabilities if instead of just functor, you pass a monad, or something that supports .filter.
Also see the possibilities of generic transformations. Every language that supports Promise needs some way to turn a list of Promises into a Promise holding a list, and vice versa. We can do this with options, or Eithers, or some home-made thing that is holding on to any other property we like... but it's so much easier to do so without having to write every possible combination of two things, and rely on the fact that one might be a monoid, or a monad, and not have to write all the boilerplate.
In languages that take this very seriously, you can do things like, say, apply a tracing library to an http request, or a logger, or manage creation of resources, all throught category theory concepts, and have all the wiring basically disappear, instead of havign to either pass a million parameters, or relying on some dubious Aspect-oriented-programming instrumentation that might or might not work. All visible, and checkable at compile time. And you get there by making all your abstractions are functors, monoids, monads and such.
It's not that they lost all of them, but that they make far less money, while the top performers in NYC do well.
It's not unlike what happened to soccer as television got cheaper and cheaper: You can go watch your town's third division team, or you can watch Real Madrid play on TV. In 3rd division nobody can be professional, in 2nd division you make less money accounting for inflation than 30 years ago, but the top players in the top teams are even bigger stars, now that the entire world can watch them play every game.
And on theater, let's not forget that many parts of the spectacle are almost impossible to take on the road. You aren't going to feed a production of Phantom of the Opera in a small town for 3 months: National tours rely on 2 weeks per large-ish metro. And when you are only going to stay there for 2 weeks, there are things you just can't get away with, economically speaking. The equity Hadestown tour would need to remodel way too much to accoout for the lift on broadway. The non-equity tour, which plays even shorter windows, can't even rely on the turntable on the floor. The car in Back to the future isn't going to fly over the audience, do half as much movement, or get fire effects on the scenario.
And even if you look in Broadway itself, many don't recoup their own costs. For every Hamilton or Lion king there are many shows that don't last 6 months.
When your company gets even a little big, the decision making process gets filtered through sufficient levels of management that it's not the company owner firing people at any time: It's an employee who doesn't necessarily have to be aligned at all with what is good for the company that is firing people at any time.
Eventually you learn that one of your middle managers managed to fire someone for some reason that is illegal, or is related to some kind of crime, and guess what? It seeps upward, and your company is in the wrong.
A process doesn't just protect the employee, it protects you from the iffy middle management that, without exception, gets in. And the more freedom you give them, the worse the behavior.
And don't forget that the percentages are not global, but in small buckets. This makes the worst performers extremely valuable, because not only you have someone to get rid of, but if they are bad enough, the rest of the team knows who will be laid off, so they can be far less tense.
It's also bad for the high performers, as working in the same team is bad: Having 3 great performers in a team means at least one, if not two are going to get a middling review. Everyone's behavior gets warped in ways that don't align well with what would be good for the company
This story of his would have fir just fine in a much smaller Stripe, back when you could actually know everyone, and Patrick interviewed every programmer. A time where making a difference wasn't all that hard, as everyone fit in one cafeteria in the Mission.
It was always a pretty competitive place, with a lot of smart people working a lot of hours, and a culture of attention to detail that left many people with impostor syndrome. There were pretty good expectations of being nice to each other: No infrastructure team giving your request the cold shoulder, because that was just not OK. So people working really hard and burning out to try to meet every growing expectations was common.
The post also had the other weakness of the culture: A lot of management changes, along with a culture of performance among managers that would be fit for Amazon. So a manager might change teams, and the person that used to get exceed expectations would end up with a PIP with the next manager, often by surprise. You can imagine what it does to morale to tell someone how they are the most helpful person they've worked with, and then see them gone two weeks later. It was a great place to work in many ways, but the negative parts took their toll.
So, if anything, the story showed me that even though it's been many years, a lot of Stripe is still recognizable.
A company making money off of this kind of scheme would be happy to pay $200 a seat for an unlimited license. And I would not be surprised if there were many other very profitable use cases that make $200 per month seem like a bargain.