fardo

> "I know some employees are going to try to steal from the register, but why would they admit that in an interview?“

A lot of the time, it’s an unfamiliarity with “test taking as being a test of determining theory of mind of the test-giver”.

Rather than answering however they think the test-giver would want the answers to look like, they approach the test with alternative less useful approaches, ones like “holistic reframed hypotheticals” where they’re thinking of the situation in a modified hypothetical with a real world context introduced and with information that might dramatically change one’s answer, such as

“assume you’ve learned a coworker has taken some money out of the cash-drawer, how much money would be your threshold to snitch on them to management versus E.G. just asking them to put the money back?”, for which many people’s answer will not be “even one dollar”.

It is a trained skill to know this is a flawed approach and that broadly, “most tests don’t want you to make outside assumptions or reframe situations, and you should default to using strict and as-literal-as-possible interpretations of questions”.

> [Being “non-technical” is a] bucket that includes knowledge of financial systems, laws, business models, operations, ethics, research tactics, user behavior, cultural patterns, learning development, communication practices, organizational psychology, and so much more

When someone has unique and actionable skills and training at any of the above, employers typically assign them a fitting and specific job title, often in a dedicated role indicating that expertise and which pays accordingly — perhaps accountant, lawyer, researcher, operations specialist, and so on.

Most of the time, if you’re role is being a “non-technical” manager you are presumed by your employer as lacking and generally not being required to possess any such specific expertise, and thus, lower pay for lower expectations.

Unfortunately for the author, the typical assessment being harsh doesn’t make the typical assessment incorrect.

Most innovation in the homeless “housing” space has focused on relocation, both benevolent (trying to reconnect them with family members and friends willing to house them and get them off the street) and self-interested (without regard to the well-being of the individual, just focused on getting them on a bus going somewhere far away).

When it works, it’s actually a great boon for the individuals at question, as well as dramatically out-competes alternatives on price, as there’s a four orders of magnitude difference in price between the cost of buying and building housing and shelters, particularly in some of America’s most expensive and overzoned real-estate markets, versus buying them a ride home to somewhere they’re expected. [1]

That said, these programs have mixed results - sometimes great and life-changing, sometimes ruinsome due to poor vetting of the opposite side and being used as a high-pressure removal tactic [2] - but cities have broadly embraced doing it because in either case, it is effective at the political goal of moving one’s homeless problem somewhere else.

[1] https://www.sfchronicle.com/sf/article/sf-homeless-bus-homew...

[2] https://www.theguardian.com/us-news/ng-interactive/2017/dec/...

It’s shocking how streaming service prices have been skyrocketing, when the actual libraries they each provide have been dramatically cut back.

I think if these firms continue recreating the old miseries of cable TV, another golden age of piracy is inevitable.

I think the article's lack of attention to both these interfaces being conversational as well as auto-translating unfortunate, as I think both of these things would suggest on average they may be substantially more accessible than alternative interfaces for the users they're calling out.

Admittedly without numbers to bear this out, I've been consistently impressed by the quality of modern text-to-speech audio transcription, and LLMs are, when provided with written dialogue, fairly skilled at replying with usable-to-spectacular results. This suggests a serious potential for improved usability for these low literacy users by expressing their queries as a verbal and untyped stream-of-consciousness by using transcription, while still giving them back well-formed results in response.

Additionally, to the extent that this is a problem for non-native speakers separate from native speakers specifically, one of the benefits of LLMs is that they often are capable of generating well-formed inputs in multiple languages trained in their corpus - this suggests that even on an interface designed without localization in mind, Prompt-driven UX can potentially give you user-language localization essentially for free.

Given these two points, I don't like the assertion that these kinds of interfaces a priori hurt usability, because they assume a framework of interaction that first doesn't have to be used, and second ignores the users whose usability would potentially be greatly empowered by these kinds of interfaces.

What’s nice for dour ending enjoyers is you can stop watching in Granite State if you want that cabin ending, assuming you’re willing to ignore the last 5 minutes or stop at episode end with the assumption being the episode ending is a call out of Walt’s hubris never letting him give up as a “Bolivian Army” ending before his demise. [1]

The show still works thematically like that, BCS largely doesn’t really care about how Walt croaks, and El Camino works equally well with “and Jesse tried to escape again, but this time got away with it”.

[1] https://tvtropes.org/pmwiki/pmwiki.php/Main/BolivianArmyEndi...

Loved your Hapland games as a kid.

Besides the standard callouts of “it’s an anomalously hard job market right now due to a recent layoffs flooding dev supply of interviewees, due to a post-covid hiring frenzy hangover, due to anticipated US tax code changes discouraging hiring, and due to high interest rates in the US currently discouraging risks including hiring” the thing that may be actually in your control is I notice is a lack of quantification of project deliverables and accomplishments on individual projects.

Looking at the first point as one example, it could be summarized “created app”. Especially as your most recent work, I’d want to see you go more into the experiences the app enabled - perhaps the user counts who were using it, what the experience was for them, any metrics you might have that you moved the needle on, or what your changes enabled people at the installation to do. Quantifying the impact of your work with each work item I think would go a long way towards making the CV better.

In my experience, this mostly is because the “hiring bar” at big companies is frequently disassociated from actual day-to-day work. It typically looks like

1. memorizing some 50-200 or so domain-specific general patterns which problem solutions will tend to follow, so you can solve an arbitrary leetcode medium in under 20 minutes time — a skill that has almost nothing to do with the plumbing and product work most tech-workers engage in on a day to day basis, and

2. Speak at length and extemporaneously on arbitrary from-scratch system design problems — a skill which, while sometimes important, is a skill which most engineers that aren’t working at a startup won’t usually exercise more than once or twice a year, as the ratio of work on “systems already designed” to “systems needing new designs” will overwhelmingly skew towards pre-existence.

When your hiring process and day job are so thoroughly dissociated and unrelated to each other, it’s not a surprise current workers who aren’t actively studying for a job change couldn’t pass it.

Cookie clicker does something similar for importing and exporting saves, and it is an absolute joy to not have to fiddle with accounts and be able to backup your own data. I’d love to see more applications move to this model, it’s very user-friendly.

> Can't folks decide that the price is too high?

If you look at drug addicts as a group, the answer is often a direct and emphatic “no, they can’t” - the drug alters the way they perform that price calculus such that many of them will spend their last dollar, go deeply in debt, and do terrible things to continually chase a high. Drug policy has had many failures in the US, but there does seem to be a correct nugget of recognition that “some drugs seem to be fundamentally destructive to those that get addicted, and to the broader society that has to deal with the consequences of those who are addicted”.

Gambling in many people seems to have the same effect on altering price calculus in a way similar to addictive drugs, which at leasts motivates a reasonable discussion whether benefits of allowing it exceed the harms it causes those individuals and society.

I think it’s more telling to look at the opposite question, “what leverage would employees have to turn such a term down, assuming such practices were considered ‘industry standard’ and all employers of note made sure to include it?”

The answer in an environment with a ban on non-competes is “they’d either have legal recourse if someone tried to write them a contract with that term, or the term would be treated as non-enforceable and therefore harmless”.

In the current environment, broadly speaking, most employees wouldn’t have leverage to turn such a term down, and generally don’t.

> Why is it actually winner-take-all?

The power of defaults, mostly.

The average user experience of picking up an internet-connected device has been very intentionally cultivated by Google. Whether you're in your browser or on your phone, Google's spent a lot of money building up Chrome as a browser ecosystem, Android on mobile, and paying off Apple on iPhones and competing browser vendors like Firefox, to guarantee that, whenever possible, Google is always the default search engine. The only non-Google default will typically be on Edge, which only has about 5-6% penetration. Since Google historically has always been the best search engine in the space, does not explicitly charge its users money, and (at least for average users) is really good at surfacing what they're looking for, most users feel no need to look elsewhere for a search engine because the default just works, switching would demand an effort, and Google is what they'd want anyways. The moat isn't big, but Google has put a ton of work into ensuing that any competing search engine requires an intentional and active choice of users to seek you out while they're worse.

At least until the recent AI play by Bing, this tiny moat was always sufficient, because if you start from scratch on search, you're essentially guaranteed to be worse, and all other 'serious' offerings under the hood were weak alternatives: essentially one of "Bing search API wrappers" (worse results), "nation-state-actor search engines" (for most users, worse results), or "Google, but with some cursory privacy measures, a subscription fee, or filtration features" (which wasn't something most users care about).

Recent chat AI represents a competing alternative to doing a search in the first place, which jeopardizes the "we have essentially all defaults and users can't be assed to switch to a worse search" barrier to entry that Google historically relies on, which is ringing alarm bells for them.

the steelman is that once you have AGI, we don’t know what happens next.

With other technology, you typically get some breathing room before problems arise - with AGI, pretty much everyone seems to agree step two once you have it is going to end up being

> someone spins up as many copies as is feasible and has them work on their smarter successor, then have those successors spin up as many copies as possible building better successors.

We’ve only ever seen human speed progress, subject to human speed limitations and human constraints. It’s possible that after a few successors the machines are much, much smarter than us, such that scary tasks we’d find challenging and technical (creating cripplingly damaging computer viruses, sabotaging infrastructure, creating life-like or real-time simulated digital communications of people you care about) could be trivial.

We don’t know how fast we go from “it’s equally smart as us” To “humans are no longer able to intellectually compete at most, if any, domains.”

Is that a problem? Maybe! Digital entities significantly smarter than humans could mean a true end to scarcity, a solution to basically any solvable technical problem we can enumerate, and a new glorious age for mankind. It could also be absolute destruction at the hands of an adversary that essentially is immortal, unkillable, of perfect memory, and who has access to the capability to gain control over essentially everything connected to a digital network, which rounds up to “everything”.

The problem seems to be

> it depends on what you’ve designed such an entity to do or what such an entity would WANT to do

And that’s scary because we don’t have any precedents for what happens when humans are confronted by significantly intellectually more powerful entities than our selves - besides how we treat animals, plants, insects and other life we see as intellectually inferior.

Judged by our past track record, assuming machines can establish their own actions with intention in basically the same way humans do, the interaction isn’t likely to be amenable to human thriving rather than machine thriving, and that means life looking more likely like “you’re the machine’s factory-farmed chicken, or if you’re lucky, its pet” rather than “the machine dotes on you and brings you into a new gilded age of plenty.”

When reading teardowns of Effective Altruism, I always find them light on substance and preferable alternatives.

I see two typical failure modes. Half of the time, they fall into a

> cannot separate the man from the idea

trap of “someone with a corrupted moral compass once believed this, ergo the idea itself is bad”, which doesn’t follow.

The other half does engage with the idea, but seemingly only ever half of it. They tend make the first step successfully,

> People can use a veneer of charity and rationality to ignore ethical issues

Which is definitely true, but don’t seem to actually take on the ideas purported main thrust, that “if one intends to give charitably, one ought to bias their choices towards causes that give money as effectively as possible at meeting your goals”, which frankly seems pretty air-tight

I think the core is that a lot of people have a problem with admitting someone they dislike had a pretty sensible idea.

> Both of these explanations strike me as too clever by half. I think the parsimonious explanation is that people are actually concerned about the dangers of AI

This rings hollow when these companies don’t seem to practice what they preach, and start by setting an example - they don’t halt research and cut the funding for development of their own AIs in-house.

If you believe that there’s X-Risk of AI research, there’s no reason to think it wouldn’t come from your own firm’s labs developing these AIs too.

Continuing development while telling others they need to pause seems to make “I want you to be paused while I blaze ahead” far more parsimonious than “these companies are actually scared about humanity’s future” - they won’t put their money where their mouth is to prove it.

Teams avoiding security reviews are common cross-industry.

Incentives everywhere are “dodge” - the expected outcome for managers and engineers is missed deadlines and reschedulings on prior promises, frequent blame-focused meetings and late nights; and low prestige, high effort work unsuitable for making the company money, getting promotions, or maintaining agility in the future, as most security work is extremely mercurial and often only very marginally improves security, if at all.

Most “security fixes” at large firms are built on inscrutable black-box internal tooling which becomes a massive single point of failure and is rarely as hardened as implied, gets re-shuffled on what constitutes “best practices” every 12 months such that a feature team can spend literally all their time doing security work the team will be throwing out next year, are often not intelligently scoped to handle “this issue doesn’t specifically apply to my component”, and aren’t even the weak link when most hacks are

> An insider replied to the wrong email, or put a zip file into a thumb drive then ran off with everything his computer had downloaded, which was a lot.

Most teams grit their teeth and implement, but if the benefit isn’t there, just the requirement, teams remember and get a lot slower to pick up the phone.

Two big things. First, because the economics of modern pinball are brutal, venues mostly stopped buying, so manufacturers and their competition shrivelled up and with it, supply. The supply-side market at this point is effectively dominated by individuals and collectors, neither who want to part with their machines, as they’re nostalgic items and they know they have value.

Furthermore, working machines come at big premiums as repairs are finicky, expensive, and time consuming - god help you if you own a machine and some targets or rollovers on your machine give out, as it’s challenging to find someone who’s even capable of doing the repairs, let alone pay their price. The expertise to do it is really thin these days because all the people who historically used to make a living from it are all getting old and retiring. This combines with a second problem: although widespread 3D printing helped a little with making repairs more doable, parts are hard to come by and often demand cannibalizing other old or dead machines - meaning a working machine is worth its weight in gold, and so is a dead one: Either it has interested players, or it’s of value for salvage. This “literally every machine is valuable” situation drives prices up, even if the machine is essentially dead.

You’re correct regarding historical procedure, but with regards to the privacy rule, which was added shortly after its creation and at least online is much of why the act is known and discussed today, the rule exists to, quoting the government’s description,

> The Rule requires appropriate safeguards to protect the privacy of protected health information and sets limits and conditions on the uses and disclosures that may be made of such information without an individual’s authorization.

We allow a major hole here in that protected health information by willfully careful readings of “appropriate safeguards” and “limits and conditions”, essentially because doing otherwise would be a nightmarish expense and pain.

If we’re discussing

> What is the motivation behind keeping medical records confidential, why do we actually care?

A respect for the patient’s privacy is likely going to be one of the driving reasons, if not the primary reason itself.

The second comment feels closer to the mark. While post-hoc justifications could be made as to why a rule at least in spirit seemingly about patient privacy ignores an obvious and glaring privacy flaw, if the parties involved could be so honest, the real-world answer why it’s allowed would probably be

> “It would be extraordinarily inconvenient and expensive for it to work otherwise.”

Sprinkle on a little bureaucrat-ese and post-hoc justification and you get the “clarified guidance” the primary comment calls out

> but then lost interest when it refused to do the same thing for boobs.

There’s a cottage industry of unaligned local LLM and image models from company leaks or released retrained public models dedicated specifically around letting you do this for “boobs” (or anything else, for that matter).

It’s not hard to find these communities if you go looking (searching for E.G “local models general” will get you in the right direction), but be advised while they’re a great resource, they’re unabashedly NSFW and unabashedly boundary-pushing, as the subset of the populace dedicated to generating their own porn and distributing porn-related models tend to be uniquely dedicated individuals.

Even if it was net neutral on hard compensation, guaranteed money now beats potential money later. Seems like a net win for negotiation.

That seems to be their theoretical use, much like the downvote is theoretically supposed to be their inverse.

In practice though, it seems that the users of most online communities seem to treat them as an "agree or disagree" button, with dissenting opinions typically being either forced out of a thread or languishing due to the upvote-downvote tug of war generally suppressing a comment's rankings.

It often quite literally will.

Free quotes and inspections prior to work commencing, time potentially spent waiting for jobs to come in in which customer support and taking calls are essentially a pure unpaid cost center for the business, taking the time to answer customer questions wholly unrelated to the job potentially even after the job completed, a desire by customers to hear you say things like

> hey, I saw your valve on blah was loose when it shouldn’t be, since I was there and it was an easy fix, tightened it for you as part of the job, no charge.

And there are likely more, I’m not a plumber.

If you are looking for it, you’ll find many cases of “free work is used as a tool to distinguish oneself, delight, and lubricate business”, and plumbing is no exception.

> I don't ask my plumber to show me their blog posts.

Yes, but you probably read his reviews online, compared his quotes with competitors in the space to pick one with a good reputation, and otherwise sought signals of competence. Blog posts are the very same thing, quickly verifiable, albeit potentially wrong, signals of skill and competence.

> Just hire people that can do the job you're paying them for

Speaking bluntly, this is what hiring people that can do the job looks like. Attacking those that take the effort to show they have that experience because others wouldn’t is tall poppy syndrome, and it neither improves the quality of the field nor actually helps those struggling that need more support so they can operate at the same level as others.

When you make stuff the order is usually the other way around - someone finds your work, then gets interested in the person behind it.

If you make it easy for someone doing so to get in contact about job opportunities and make it clear you’re open to them, you’ll often start to get some.

If possible, redesigns preceded by targeted warning emails sent to business customers based on message length stats, with a timetable for when the fix will rollout.

If that data doesn’t exist, public announcement that the API is breaking long messages without the new parameter.

If the API can’t be broken, version the API and break it in the new one and for new customers.

In all cases, disclosure in the documentation - If you know there’s a limit and you still exceed it, at that point, as the kids say, FAFO.

Defaults are powerful, and IE was default.

If you’re a large bureaucracy in the early 2000s you went with whatever was default - and that was IE.

Defaults were seen as so powerful, that MS was anti-trusted for this. They did it anyways at the time because the juice was worth the squeeze.

Pay them or form partnerships with companies to either assign part-time or full-time developers to them, or pay contributors.

Open source is attractive in major part to companies because it’s externally vettable and you typically get a stronger ecosystem of enthusiasts working on your software by using it. Open source projects that has consistent enthusiast development has some leverage to negotiate deals with consuming companies to keep that ecosystem strong.

I wonder how much of this change is driven by Russia’s failings to achieve objectives in their ongoing Ukrainian war.

When another actor is perceived as a peer military, you have strong incentive to have closer relationships with them so you can influence their behavior. If China’s ability to win wars against neighbors is perceived as having gone down due to Russia’s struggling, much of the need for economic influence would seemingly go down in kind, allowing priorities to shift.