fardo

This feels like it should be a non-controversy.

Even being uncharitable, a big off-by-default checkbox saying “make this discoverable in web searches” is roughly as explicit as you can possibly make this feature textually, assuming your users will be applying any reading comprehension.

If they’re not, no further warnings were going to save them, so short of removing the feature or gating it behind increasingly elaborate “if only you knew better!” emails or pop-up modals they also presumably would not be reading, this was the likely outcome.

At some point, I don’t feel bad saying this is a user-side PEBKAC, and that more alerting would be a waste of time.

I assume that he means "rather than pushing up each individual container for a project, it could take something like a compose file over a list of underlying containers, and push them all up to the endpoint."

They have a lot of ways they could’ve built trust without a full negative burden: which of them, if any, are they doing?

Open sourcing of their watch word and recording features specifically, so people can self-verify it does what it says and that it’s not doing sketchy things?

Hardware lights such that any record functionality past the watch words is visible and verifiable by the end user and it can’t record when not lit?

Local streaming and auditable downloads of the last N hours of input as heard by amazon after watchwords, so you can check for misrecordings and also compare “intended usage” times to observed times, such that you can see that you and Amazon get the same stuff?

If you really wanna go all out, putting in their TOS protections like explicit no-train permissions on passing utterances without intent, or adding an SLA into their subscription to refund subscription and legal costs and to provide explicit legal cause of action, if they were recording when they said they weren’t?

If you explicitly want to promote trust, there are actually a ton of ways to do it, one of them isn’t “remove even more of your existing privacy guardrails”.

Aspersions aside, the content’s actually typically pretty involved and has a lot to speak for itself, it’s not low-effort content that one would typically associate with AI.

Laid bare, it’s generally a variety comedy show of a human host and AI riffing off each other, the AI and the chat arguing and counter-roasting each other with human mediation to either double down or steer discussions or roasts in more interesting directions, a platform for guest interviews and collaborations with other streamers, and a showcase of AI bots which were coded up by the stream’s creator to play a surprising variety of games. There’s a lot to like, and you don’t need to be on “that bit of the bell curve” to enjoy a skilled entertainer putting new tools to enjoyable use.

I think this optimism is extremely misplaced, as I think things are likely to get substantially less evenly distributed. It seems more likely we'll have a future where indies more successfully drown each other out in a sea of noise, while major studios continue to enjoy massive moats of content and marketing.

On the content front, studios have benefited massively by selling sure bets in a sea of noise: of the top 10 box office films of 2024, there isn't a single original IP, every single film is either an adaptation or sequel to prior work made long ago. [1]. I view this as part of a broader "flight towards quality" pattern in the internet age - even if there's tons of great content online (orders of magnitude more), the viewing public still ultimately values both studio curation and IP familiarity. This goes beyond the films themselves: the studio IP rights moat includes the access to famous big-name actors that fill seats, the IP rights to use their likeness, and access to their personal platforms to push the film, all of which these studios control. Even if indies can generate "a person" or "a movie", the inability to legally generate "specific people that the public knows and likes without their permission" or "specific movies set in universes they know and care about" represents moat that isn't leaving studio's hands in a world of widespread AI.

Separate from IP rights, it also assumes a hyper-specific model of how AI specialization and use will look if it's widely available. For example, even assuming AI that can generate anything you ask for, it's likely that people will continue valuing significant sound, music, and visual post-processing to augment end-state AI generations to better match and personalize a final vision differentiated from the models themselves. This means labor costs, which indies at scale would continue to lack access to. This also assumes a model of the world where AI becomes a reducer of specialization, which isn't guaranteed: even assuming superhuman production capabilities by AI, such that there is no longer any individual human input on some aspect of the film, someone has to point the superhuman AI in productive directions that map to somewhere good in the quality spectrum, and it's likely that there will be significant differentiation in human skill at this task across different domains, as one can think of the superhuman AI as a tool being consumed by a skilled collaborator, even if the AI is doing most of the work. In the event that this is the case, studios can and will continue having much bigger labor budgets to continue to differentiate themselves on quality compared to indies trying to DIY this process.

And lastly, marketing is still king. About half of a modern big-budget picture's budget gets spent on the marketing today, only the other half goes towards making the actual film. Even if state-of-the-art Hollywood grade AI means anyone can produce a shot-for-shot reproduction of any current Hollywood film, ultimately, people need to find the content in order to watch it, and even in a world with widely available AI, it's the indie's marketing budget of ~$0 dollars, versus the major studio's marketing budget of $50,000,000-$200,000,000. I would happily continue betting against indies winning this fight, especially when the low end of this market, which is already oversaturated and hard to meaningfully stand out in, is drowned out in an order of magnitude more noise from low-end AI creators flooding the space with slop.

[1] https://www.boxofficemojo.com/year/world/2024/

One of the best sonic games ever made, sonic mania, wasn’t made in-house by sonic team, but a handful of fans making fanwork.

Sega’s choice to treat their fanwork seriously and make them second-party developers rather than shutting them down was not only an immeasurably good thing for both the project and its fans, but a measurably, handsomely profitable move for Sega.

The author’s right about storytelling from day one, but then immediately throws cold water on the idea by saying it would have been a bad fit for this project.

This feels in error, as the big value of seeking feedback and results early and often on a project is that it forces you to confront whether you’re going to want or be able to tell stories in the space at all. It also gives you a chance to re-kindle waning interests, get feedback on your project by others, and avoid ratholing into something for about 5 years without having to engage with a public.

If a project can’t emotionally bear day one scrutiny, it’s unlikely to fare better five years later when you’ve got a lot of emotions about incompleteness and the feeling your work isn’t relevant anymore tied up in the project.

It shouldn’t be surprising that a country with a war economy has a higher first derivative at producing material, the question of import is

1. What is the difference between absolute quantities comparing against all relevant players,

2. How long would it take to bridge the gap at current production rates, and

3. Can that rate of production be sustained long enough for it to alter any fundamentals?

The point to rebut isn’t that Russia is making more, it’s whether they can continue to do so ongoingly before Ukrainian advances, regime falter or economic collapse, US/China step-in, or internal unrest will dramatically weaken or make the current Russian negotiating position untenable.

In an era of massive scale companies and giant projects, it's fascinating how often it feels like most of the success of a project or company's big successes ultimately hinge on the actions of a few key individuals in the right place at the right time - and it never fails to surprise me how little overlap these individuals and the actual actual company org charts share.

What is the problem with long lived certs?

Privilege escalation and Dev Ops rot. Long-lived certs often get compromised when privilege escalations happen and someone gets access to an account or computer that has private keys on it.

One example scenario for privilege escalation: let's say a hacker gets access to one of your employee's or vendor's machines and associated accounts using a zero-day, or phishing, or some other method that goes undetected for some time. The attacker, as part of this attack, successfully gets access to your cert's private keys through some way or another without drawing attention to themselves.

Some time later, your firm makes several security updates. When doing this, you unknowingly patched the attacker out of your network. The attacker is now in a race against time if they want to do something with the cert before it expires, and in this kind of situation, the sooner that cert expires, the better, because the attacker gets less time to do something with it. In a perfect world, the cert expired exactly when they got patched out, but because we're not guaranteed to know if there's an attacker in the first place, "keeping the expiration time as short as is reasonably possible without impacting service reliability" is what things seem to be moving towards, to limit the blast radius during access leaks.

As for Dev Ops rot, speed has a tendency to change requirements in favor of automation. Generally, certificate rotations tend to be a pain point - they break management panes, they take down websites, they throw browser errors, they don't get updated in pipelines, and other woes happen when they expire that demand people keep track of a ton of localized knowledge and deadlines that's easy to lose or forget. However, paradoxically, the longer the time between rotations, the more painful they tend to be, because once rotations are sufficiently fast, it becomes unmanageable to do them manually: demanding speed forces people to build anti-fragile rotation systems. Making the requirement be shorter is in some sense an attempt to encode into managerial culture "you need to automate this", as a bulwark against swapping your certs out being anything besides automated or one click rotations.

define more competitive markets - Is it even possible?

Yes it is, and I am using it in the traditional sense of "Sellers vying to make their offerings more compelling so consumers prefer them to alternatives". A more competitive market has more of the above fighting happening, which can be incentivized in several ways.

Specifically for the US, I already gave a massive truck-sized example of how you can make offerings more competitive: repeal recent 1970s policy changes that make college debt work different than essentially all other debt in being non-dischargeable, and lenders will put pressure by refusing to lend for failschools. This will slash their demand, and survivors will be expected to provide more compelling “when examined actuarially” lifetime student economic outcomes: from the lender’s perspective, the education needs to be expected to at least pay for itself. Schools in a world with dischargeability will therefore make efforts to compete on being able to compellingly show “we are imparting an effective education which equips you to repay what you spent”, and in general will be less able to get away with “we have a strong brand and you’re young and naïve, come study here!”. You can fool an 18 year old, you won’t fool the underwriter.

Specifically for the EU, there’s three approaches I’d like to see considered, the first and last more experimental than the others. It’s a widely agreed upon problem that faculty aren’t incentivized or often even that good at actually teaching, and to a lesser extent they’re also bad at training graduate students, because professorial incentives are to be good at research and grant writing, the thing for which universities pay their salary.

Historically, many universities used a different mode where they kept permanent salaried faculty positions very limited, and rather than charge a tuition, most schools and professors were expected to charge a fee or honorarium they would collect from their individual pupils as the bulk or totality of a professor’s compensation.

We moved away from this because historically teacher compensation was abysmal and we wanted to better support educators in studying - the pendulum swung way too far the other way though, and professors are essentially totally insulated from needing to seriously engage with their student body - they make a salary no matter what and it shows when they teach like it. Returning to a market where teachers’ financial outcomes are at least partially linked to their students explicitly choosing and paying professors chosen fees rather than a tuition which everyone in society pays stochastically, professors are incentivized to competitively vie for being better at training and teaching, because their prosperity and success would be linked to it.

Second, as much criticism as the US model gets for its prices, much of that appears to be driven by administrative bloat and nondischargeability, payment itself is actually really great when dischargeable because it seems prices stay low and this bloat doesn’t happen when it is. It also means a significant fraction of students do try to get into the best school they can and the schools have some incentive to fight to be seen as better - they make their money when students go. The results speak for themselves: something like half to two-thirds of the world’s top 100 universities are all in America.

Last, I’d like to see both European and US models collaborate harder on job and work placement as part of ensuring student success. Colleges being a source of a liberal education without concern for fiscal outcomes can be great, but the pendulum swung too far and these students are essentially thrown out into the world without any expectation of success, because the school no longer has any investment in them. I’d like to see some portion of the fee of education be moved to explicitly garnished wages from placed-into-their-trained field employees - you succeeding would therefore be the university’s success.

I’d therefore like to see taxes slashed on education with moves to paid models comparable to how they existed in the US prior to the 70s, reductions in permanent salaried academic positions with in-kind movements of price of tuition from the school at large to being in baked into professorial fees instead, and moving the funding of education from societal fees to a fixed percentage of post-graduation salaries for some fixed time, to incentivize effective job placement.

The common bailiwick that needs to die is that Europeans don’t pay for college - they do, it’s just compulsory, lifelong, and in a way that is not incentivized to produce quality.

has it ever existed?

This question being written as singular suggests to me you’re conflating some combination of education as a concept, education in America, and education in Europe. I see this as muddled thinking, they each have their own issues and their own timetables over which issues can and do exist.

Education history in Europe is actually decently hyperlocal and pretty path dependent for each country during the 19th and 20th century. Talking about it as a monolith is challenging, but there’s a pretty consistent pattern in societies with all-pay taxation-paid schooling that it has clear incentive structures that a school, having been “paid up front“, is not incentivized to put in the work to provide a high quality education, and as the financier of that education is the taxpayer, the pressure is primarily to keep costs low.

As for America, you don’t need much faith to believe that in the US instituting dischargeability would dramatically improve competition and quality in these spaces and the market for education - so yes, this market used to be substantially more competitive and used to exist. Since the 1970s when these laws were passed, student debt and tuition costs have skyrocketed, low quality for-profit university creation skyrocketed, and people have been hobbled with debt for degrees many are never going to pay back. Dischargeability reduces this greatly - it causes financiers to create motive pressure on education to not be a fiscal vampire preying on its students and equip them with meaningful trades and socially useful knowledge.

As for standards and regulations on schools, assuming the public sees fit to fund higher education through government spending, having requirements for a school to receive that government funding seems reasonable, so long as they reflect the desires of the local populace.

Ultimately, all this is a discussion of tradeoffs - I think someone can happily prefer

everyone gets a free tertiary education paid for by their taxes

I think that it being free and paid-by-all means it’ll have low motive pressure towards quality and therefore be comparably mediocre to free elementary and secondary compulsory education, which even based on your own reckoning seems to be a two-tier system where the public institutions get short shrift - but you seem to care about it being free or less focused on fiscal student outcomes as its own value. It’s a tradeoff.

This is non-responsive to my point. Pointing out that paid universities like Corinthian in a grossly distorted market predictably are not very competitive offerings, or that good offerings can nevertheless still exist in heavily distorted markets, both fail to address that

Current educational incentives caused by how payment is handled in these all-pay systems mean there is very little or no pressure exerted towards promoting an educational arms race towards quality, rather than minimizing cost to service that education.

In a more competitive market, yes, I believe you'd both likely see better European offerings, as well as substantially more compelling American paid offerings than the current batch of cash grab for-profit universities. They exist in their current form almost exclusively because financial lenders in the US have no incentive not to issue loans to students - even if the program is bogus, they are guaranteed re-payment.

This is a commonality shared between both current European offerings (via taxation guaranteeing repayment), and American alternatives (via guaranteed load repayment).

Simpler isn’t always preferable: note that the key feature from the consumer perspective in that system is

you pay no matter what

Meaning if it’s assumed “despite learning little, you still will be able to pay for it”, there’s no longer any motive force towards quality, as your payment is assured. Incentives on taxpayers thereafter who want to minimize their tax burden would therefore be optimizing primarily toward cheapness of the educational process, rather than efficacy and quality of the education which outbound students received.

A vigorous market for education, would suggest you would likely get a variety of nodes along a “costliness to quality” options frontier.

Nice, this problem is isomorphic to the probability/graph theory problem “hunters and rabbits” from Matousek’s discrete math textbook, except with the slight modification that instead of “n hunters with perfect accuracy each randomly and simultaneously shoot one rabbit among a set of m rabbits”, it’s “n birthday havers each simultaneously have a birthday among m=365 days”

There is a closed form solution to this problem’s expectation for arbitrary n and m, which I’ve linked below:

https://math.stackexchange.com/questions/610250/a-question-o...

First of all, we're talking about pre-IPO startups, so you can't just sell the stock

This is often but not universally true - about 40% of companies allow you to do so, and about 40% of those that allow you to do so allow those sales on secondary markets [1]

company loses money because it gave equity to employees is nonsensical

You lose money in the sense that the gains beyond the strike price which you would have realized had you not sold the call option are your losses: you could’ve not sold the option and profited on that rise instead. You have lost the difference in the profit between these two investment plans.

You're not giving away equity

I was responding to someone asking “why don’t they just give the equity instead as compensation?”, as opposed to writing call options against it - assuming that as a company you want to incentivize workers to work by setting aside some fraction of your equity which they may receive, these are the reasons a company might prefer to “give” employees that equity with options, instead of discounted equity or stock grants

[1] See page 8 footer of https://www.gsb.stanford.edu/sites/default/files/publication...

Actual equity is hard cash, options are a potentiality of cash.

As the company with equity, any calls you’re selling are guaranteed covered, which hedges against downside loss of having given away equity and it exploding in value. Calls also theoretically align incentives better than equity because it gets the staff member personally interested in seeing the stock rise, rather than just selling immediately to take profit if they’re short your company’s future outlook.

Lots of other nice properties - If you sell the option, either discounted or at some full price, you make a money premium, which helps runway. If the stock falls, you make money from having chosen options and basically don’t have to give the employees anything. If the stock rises, you’re probably making great sums from your (much larger) equity share, and your losses versus just giving them equity are essentially capped at the difference between the current price and strike price, instead of theoretically unlimited cost. This is easily quantified with some multiplication when issuing the options, meaning in a growing company, you have knowably limited exposure, and essentially are just giving them the equity you would’ve given them anyways.

The CRT requirement has pleasantly eroded recently.

A kickstarter a few years back for the Sinden light gun [1] realized that by using webcams, some quick image processing and perspective transforms, you could make a light gun work anywhere and could get real-time performance on non-CRTs by essentially adding a small border region of the screen, making it work on essentially any monitor. He filmed and wrote extensive technical breakdowns about the build process and mechanics at play, which were great.

The maker also seems to have had a solid understanding of what made those old light gun games cool, because he made sure to build versions with solenoid-based recoil as well as the big chunky metal foot pedal you’d use for games like time crisis.

[1] https://youtu.be/grcGpr_8W9Y?si=z800V7f62dDS1KGs

Pixar badly lost its edge by getting complacent.

Somewhere after cars, 3D animation competition started breaking past pixar’s moat that no one else could make a 3D movie look quite as good, and somewhere after Toy Story 3, it became clear that competition had broken past pixar’s moat that no one could tell equally compelling narrative stories. Since then, Disney’s in-house team, Dreamworks, and other major productions have soundly eaten their lunch.

Pixar needs to return to its bread and butter story-telling, trying to focus on being a toy-peddler has ruined them.

Usually shortcuts are good for lock-in because besides being useful, they get you used to certain workflows. For example, excel has a useful shortcut where "CTRL + ;" inputs today's date, and "CTRL + SHIFT + ;" inputs the current time, which you can do in the same cell to create a date-time instantly with almost a single motion.

It was a source of endless frustration when once I was extremely used to this shortcut, I later switched to google sheets for a different project, and the shortcut didn't work the same. Google still supported both individually as they had copied excel, but the way they work on sheets is before inputting either the date or time, it would delete whatever existing value was stored in the current cell, meaning you could no longer instantly input both a date and a time to make datetimes, you instead needed two columns where before one would do.

Shortcuts increase the frequency of these kinds of friction points with your power users when they move elsewhere.

This bodes poorly for the future of SO.

One of the reasons that Quora today is absolutely unusable is that it no longer is a curated discussion between internet users and knowledgeable people, but AI spamming the site with swarms of low-quality questions, and AI answering those questions with swarms of low-quality answers. I think it's likely that Stack Overflow will end up following a similar pattern.

I would. Most modern systems of education have their roots in 1748, and either are derived from or inspired by reforms to the Prussian educational system to guarantee free and compulsory elementary school education between the ages of 5 and 14 as taught by secular professional teachers for the full populace.

https://en.wikipedia.org/wiki/Prussian_education_system

If you mean the university and college levels, they have interesting differences from the 18th century, but are recognizably similar in regards to basically everything besides cost and curriculum differences we'd expect due to changes in societal needs, technical advancements, and changing interests.

On-the-job observations done by someone with way more experience

This feels like it puts a potential hard cap on quality growth by discouraging mixups or experimentation that might improve education, but wouldn’t please an old-guard for one reason or another, and discourages alternative class styles which the judge doesn’t approve of.

Both of those seem like potentially serious problems in education, given that its structure has with few exceptions been effectively stagnant over the last several hundred years. You therefore may be mistaking “evaluating the success at implementing the widely accepted method” for an “evaluation of quality”.

I wouldn’t discount the reputational and long-term danger for the school of potentially losing “peak reach school” status, which this change may be symptomatic of.

Harvard has long been the beneficiary of a reputation of “if you get in, you’re very likely to go, as it’s probably also your top choice or very close.” This maintained an upward pressure force in Harvard to keep it a top school, because short term a strong reputation drives top-ranking students that want to study with other top students, and long term those high quality students are likely to accomplish more and further reinforce the reputation.

The flywheel starts to break if when you’re choosing between Harvard and somewhere else, you pick somewhere else. These kinds of small shifts are the rumblings you’d expect to see if the edifice was beginning to crack.

I think it's three things:

1. The hangover in tech caused by post-Covid overhiring and layoffs, as well as negative sentiment surrounding inflation over the last year feels like it's ending or at least calming down - somewhere between lower staffing spend internally and a rosier feeling economy, people feel a higher valuation makes sense.

2. Meta's riding and part of leading a general tide that's lifting all tech boats due to hype in the AI space surrounding recent LLM developments. Meta successfully releasing frontier models in the space shows they're capable of being a crucial player in the space and still have the talent internally to compete with other big players if they so choose, although thus far they've felt content with releasing free models as a broad-side attack on their competitors' moats. The market additionally wants to cost in the non-trivial possibility that these technologies will either receive massive uptake across other industries, or the possibility the tech's power and intelligence will massively improve in the next few years, potentially even to the level of something approximating or beating an average intellectual human at knowledge work, which would make all these tech companies making this technology very, very rich.

3. Focus in the company appears to be shifting back towards their core competencies of social media technologies, taking ground from Tiktok with new offshoots such as shorts with Reels, and as described above, their AI projects, which all feels like a return to form after their mostly ill-fated VR and metaverse projects, which are now being de-emphasized. This is felt as boding well for the company's future growth prospects, as those projects felt like dead-ends.

There’s a tension between the common belief that

“when private citizens are able to vote privately, it protects their ability to vote their conscience, rather than allowing some third party to explicitly buy votes or bully someone into voting in line with someone else”,

and the belief that somehow this doesn’t apply to congress members.

Additionally, on hard philosophical and policy qurstions, some bits of negotiation and dealmaking are bare-knuckled “the sausage gets made” affairs that are brutally hard on the ego and participants. Part of why nothing can get through Congress anymore in a timely fashion and without continual brinksmanship on important funding or to prevent shutdowns is because even if crossing party lines would very often be in the public’s interest, and to the public’s net benefit, haggling to make it happen or voting to make it so often doesn’t stand up to the scrutiny of thousands of watchful eyes where an important deal may hinge on brutal haggling which the public couldn’t stomach seeing the intermediate steps and votes of.

One such example in practical terms: if the constitutional convention which replaced the articles of confederation took place in the internet age with modern real-time, to the minute reporting on how everyone was voting on every intermediate plan and how any compromise made was a betrayal of “party lines” on an issue, America as a country probably wouldn’t exist today.

Transparency has its own benefits, but it’s not without costs - you make a legislative body’s job more difficult, you get corresponding gridlock to match.

The writers have only ten seconds to persuade the reader

Sometimes they get barely even get one sentence: The article cutting to a paywall after the very first line feels accidentally prescient.

I think there’s a tension between the tinkerer, blogger and hobbyist side of the field, and anonymous 9–to-5 workhorse line of business programming in this discussion.

The core reason everyone uses it seems to me to be essentially the same reason the former groups hate it - it’s what everyone else does, it’s an incredibly square and “day-job”-esque approach to one’s tooling and architecture, it often reeks of bureaucracy, compartmentalization, and organizational superstructures often unrelated to the technical work, dictated primarily by managerial fiefdoms and needs to organizationally coordinate, and it’s a well-defined space with extremely predictable solutions and headaches, and therefore has very little excitement or novelty to offer.

I feel like some of the fault rests in car authentication tech being downright antiquated by having a “whoever has keys is driver” policy.

We wouldn’t accept, in any corporate environment, a computer system where the only form of authentication was a yubikey with no password, the fact that our cars essentially still work like this in 2024 is appalling.

I think part of the disdain is people’s recognition of Google’s inability to make and maintain new products.

Separate from recent poor generations, I’d be reluctant to use any Google AI, as if for some reason I’d liked it, the odds are too high it’ll get shut down or replaced by some worse internal competitor.