md_

A potentially simpler explanation is an A/B test of a potentially CPU-intensive feature, and a misinformed data scientist.

Testing a higher-compression image format might effectively be testing relative behavior between an experiment arm with higher data usage and lower battery usage and an arm with lower data usage and higher battery usage. I.e., "testing draining users' batteries."

There are too little details in the blog post or the original article to really figure out what's being discussed.

I think the simple answer is that "investors demand it."

Investors may indeed be predicting a recession, and it's probably true that many of these firms overhired. But there's no specific fiscal reason for most of these companies to cut staff--5% the cost of their workforce is not standing in the way of billion-dollar deals (e.g. MSFT/Blizzard) or long-term R&D spending (Meta, whatever you want to say about their Metaverse project).

Bluntly: cutting staff is the new stock buyback.

> These a companies with globe spanning businesses the touch much of the world every single day (I’d put Stripe in this bucket too). They’ve got much greater and earlier visibility into the state of the broader economy long before governments do.

If that were true, I'd expect them to have hugely profitable proprietary trading desks, which does not seem to be the case?

I found https://news.stanford.edu/2022/12/05/explains-recent-tech-la... to be very insightful, framing layoffs as "social contagion."

As others noted here, irrespective of whether there is a coming recession, for companies like Google and Microsoft, 10-20k employees are not so expensive as to be existential for the company, *or even for specific opportunities the company wants to pursue*. (E.g. for Microsoft, buying Blizzard is not predicated upon laying off employees.)

Instead, it's best to see this as a sort of social trend: layoffs to demonstrate fiscal responsibility to investors, in order to promote equity valuations that are in general valuable (either strategically, or as the ultimate purpose of a publicly traded company).

If that sounds like I think this is all reasonable and justified, I don't. But it's a reasonable thing to do if your sole goal is profit, which is how we generally look at public companies in a capitalist economy.

I don't follow. Why do meat eaters want vegans to eat this instead of soy meat?

I am mostly vegetarian, and I quite like Impossible Burgers. Sadly they're not really available in Europe.

They might be an idea with a small market. But in my mind this is a lot like if Silicon Valley got really pumped up about, like, unpasteurized cheeses or nutritional shakes. (Oh, wait, that last one happened.) The problem isn't with unpasteurized cheeses--there is a small but reasonable market for such things--but with Silicon Valley.

This wasn't me not understanding why it might be valuable. This was me asking you to explain what makes you think this is valuable, how valuable you think it is, and why you think it's a unique advantage to Tesla.

One way to look at this would be the ARPU from data-intensive services like Amazon, Facebook, Google, etc. Typical orders of magnitude here is dozens of dollars per year per user; I think Amazon shopping is closer to $100 if I remember right.

I dimly recall some idiotic tech blog claiming that in the future, cars would be free because of the value of ad targeting based off of collected car data, which is hilarious:

- Cars cost three orders of magnitude more than any reasonable revenue per user from such ads

- Cost of a car per year (depreciation + maintenance) is, what, 2 orders of magnitude higher than that?

- There are cheaper ways to get that data, like...free smartphones? Yet nobody does that.

"Data is the new oil" sounds good, but it leaves out a lot of important details, like "how do you monetize that data" and "how much do you make for it?" If monetization is ads, the answer is "not much."

Presumably a combination of survivorship bias (we are most aware of the successful tech companies with asshole-founders-and-foosball-tables-in-the-lounges, and not all of the unsuccessful companies with the same) and fundamental attribution error (we think having an asshole-founder-and-foosball-tables is predictive of or causally related to being successful).

Well, Keynsian Beauty Contest, then. As long as you play it better than the others, you win.

I don't do a lot of individual stock picking, because I can't be assed to look too closely at fundamentals and I don't think I have a good enough pulse on the public opinion that I can predict who Redditors will think is a sexy CEO next.

I guess that's true. FWIW, because this chart only goes back ten years, it's not like there's some obvious trend with WalMart (as an example) where as revenue increased, margins went up: https://www.macrotrends.net/stocks/charts/WMT/walmart/profit.... In fact, as revenue went up, margins went down-ish.

The comparison I would still draw is to pure software businesses where they:

a) Pay almost nothing per unit b) Increase in value the more users they have (due to more software for the platform, for example, and greater compatibility) c) Directly lead to significant productivity/efficiency gains in their customers, which they can then claim a piece of

If a retailer becomes a near-monopolist, short of monopoly pricing, they will generate margins that tend towards the average for retail (which are low!), and they will grow at close to the rate of economic growth.

In comparison, if a software company becomes a near-monopolist, even without monopoly pricing, their margins will increase and their utility will increase.

Definitely. I'm not claiming it's black and white or either/or, to be clear.

But, what are the characteristics investors should look at for "tech company P:E"? Presumably a non-linear profit curve driven by low unit costs, linear or increasing revenue per unit, and strong network effects. IOW, if you're building the next Microsoft, each unit of Windows sold costs you nothing, makes you the same revenue (and thus more profit) than the previous unit, and increases the appeal of Windows because "everyone is using it."

Retail businesses are just totally the opposite of that.

The network effects are limited. (I guess for marketplaces this isn't totally the case: if everyone is buying on Amazon, then more resellers want to become Amazon resellers. But then, for "platforms" like Uber, it seems the network effects are weaker than we might think; drivers and users both find it easy to drive for/hail on Uber, Lyft, etc, side-by-side.)

The unit costs are fixed.

As the retailer gets bigger, their growth naturally trends closer to overall economic growth. (If you sell software, and the software makes workers 10x more productive, you can expect to get a cut of that 10x in productivity. If you sell milk, your market is going to grow at the rate at which demand for milk grows.)

The real malefactors, in my mind, are people like Warby Parker, Away, Casper, etc--direct to consumer is fine and well and probably lowers costs a bit, but it's fundamentally similar to ordering from the Sears Catalog in the 19th century. But by some bizarre combination of hype and, I know I keep saying it, clever CSS, these jokers have convinced investors they're somehow different.

See https://www.economist.com/business/2021/09/09/direct-to-cons..., https://www.ft.com/content/616421f0-6946-485a-aca4-e9a2a522a..., etc.

AWS, yes. Retail? Nah.

That's an odd definition for "tech company." Is your point, though, that that data are valuable? If so, what for?

This is true for so many "tech" darlings: Should Uber be priced like a tech company, or a cab company? Is Amazon a tech company or a retailer? Are ecommerce outlets like Warby Parker and Away tech companies, or just retailers with good CSS?

It continues to amaze me how investors seem bamboozled by the appearance of "tech company-ness", by businesses which are obviously in markets that don't support the high price to earnings of "tech companies"--valuations which are premised on strong network effects (present, to a degree, with Telsa, in the form of the charging network--but they will open that up to other carmakers, I guess?) and low unit costs (um, not for cars), and which fundamentally make no sense for things like retail (high unit costs, no network effects) or scooter rentals (high capital costs, high unit costs, limited network effects).

But investors--showing the wisdom of crowds, I guess--seem frequently unable to look past the shiny CSS and the fact that the founder/CEO wears a turtleneck and the headquarters office has a beer tap.

The idea that a social network might ask their government for assistance in identifying foreign influence operations...is pretty run of the mill?

Not sure what a "monger" is in this context.

I mean, hypothetically, what if Twitter said, "We really think it's bad for our brand to be connected to state sponsored info ops, and we'd like help avoiding it. Let's reach out to the FBI for threat intel to help us keep abreast of info ops."

(To be clear, I'm not saying this is what happened. In particular, Twitter's allowing Pentagon information operations--apparently by accident--on their platform doesn't look great.)

As a hypothetical--is that bad? It seems like it runs afoul of your "government should play no role in policing speech." But that seems like a somewhat oddly absolutist stance to me.

(For another hypothetical: let's say I'm Fox News, and I want to have a commentator on, but first I ping some buddies at the FBI and ask, "Is this commentator known to be an unregistered foreign agent?" Again, is it bad?)

A lot of these things could be bad. We might argue for some sort of firewall between anything involving media and private speech and anyone working in government ever. But that seems like a pretty radical departure from the status quo, and it's not obvious to me that what Twitter did was in bad faith or really different than the above examples.

That thread seems to show two things:

1. Trust and safety is a hard job, and Twitter was...kinda not great at it, I think? Honestly, not a big surprise.

2. In high-level discussions with government policymakers, policymakers expressed general desires for what kind of misinformation might be of particular concern (like health misinfo, or misinfo leading to hoarding and panic buying).

#2 seems...sorta fine to me? #1 seems like a product problem, to be perfectly clear, but it's hard to connect that to government censorship.

I'm not trying to make light of all the failures in #1. But I find it a bit hard, assuming good faith, to understand if the view of most critics is:

a. There should be zero content filtering or content-related ranking. (I mean, that's fine; Mastodon doesn't have algorithmic ranking, and if you don't want filtering, enjoy the Viagra spam I guess?)

b. There should be content filtering and ranking, but it should rank things the way critics want. (For example, a lot of critics seem concerned not with ranking in general, but with downranking of heterodox views on Covid.)

c. There should be content filtering and ranking, but Twitter just wasn't great at it; we'd like to see them improve their processes (which I am not optimistic about given the headcount story at Twitter).

Or something else?

Can you point me to the “Twitter Files” showing the government set the bounds of acceptable topics? Like I said above, I had trouble picking through these long tweet threads.

It's a neat theory, but Apple was 4% of Twitter's Q1 ad revenue. I don't think the My Pillow guy has $100M to spend on anything, let alone Twitter ads.

Crypto scams and fake nutrition pills aren't the top advertisements on Infowars because Alex Jones knows more about the media business than Fox or CNN. They're the top advertisements on Infowars because mainstream advertisers won't touch Jones with a ten foot pole.

Same goes for the Twitter that Musk built.

Does it turn out that Twitter was shadowbanning? As far as I could tell from the "Twitter Files", no. The only thing "revealed" was that...

...Twitter ranked tweets? Which they always said they did? (e.g. https://twitter.com/bendreyfuss/status/1601019554980761600)

Even stranger, Musk himself seems to be a big fan of this approach? ("Freedom of speech, not freedom of reach"? https://twitter.com/elonmusk/status/1593673339826212864)

For those of us with, like, day jobs, can someone summarize what damning discoveries were actually revealed in the "Twitter Files"?

My impression is:

1. These "independent journalists" like numbered lists instead of prose (OK?)

2. I find numbered lists a bit hard to read.

3. There's no headline or lede that suggests...anything? But I keep reading the numbered lists out of morbid curiosity.

4. At some point it's revealed that, shockingly, an American tech company was in contact with American government officials.

5. Those officials spoke to the tech company about known or suspected foreign influence operations that might affect that tech company.

6. The tech company expressed gratitude for those tips.

Like, honestly, isn't the reason the "mainstream media" hasn't reported on this because they're in the business of reporting on things that their viewers would find interesting, and this is a super boring story?

Yes, if you name anything "the [something] files" and label it "suppressed knowledge now come to light", it sounds sexy. But is there anything actually significant here? If so, can someone summarize it for me in a numbered list?

I think the poster isn’t talking about the remote server seeing client MACs (because, as you say) but, rather, about the way mobile OSes use nearby wifi access point SSIDs to provide GPS-less geolocation. Samy Kamkar raised my awareness of this way back in 2010 or so, and this article about his DefCon talk provides a good quick explanation: https://www.bbc.com/news/technology-10850875.

That thread is really hard to follow.

Is the claim that Twitter changed their ToS in order to justify banning Trump? If so, can you share the before and after texts? I assume the Internet Archive would have snapshots.

Or is the point, literally, that people at Twitter discussed whether a change of policy was a good idea in the context of the Jan 6 insurrection? In which case, like...wouldn't you sort of expect them to have conversations about the fitness of the ToS to an unprecedented situation? That sounds like doing their jobs competently, no?

I don’t think “how do I write differential equations in LaTeX” is a high-ad-revenue search query? Whereas a query like “what’s the best car insurance” or “where can I buy air Jordans” is.

Does GPT do a good job on those types of queries?

I suspect the challenges in running a scalable distributed realtime system like Twitter are fairly significant. And while I don't know how much of Twitter's operating expenses are attributable to the efficiency of that platform, if the new CEO wants to keep expenses lower than revenue, they would potentially be very concerned about how to make that platform efficient, which probably has some really interesting systems engineering challenges.

Maybe I'm living in a filter bubble, but I think it's going to be hard to build a "super app" that relies on the presence and active participation of reputable brands--publishers, if it's a news platform; businesses, if it's a b2c communications platform--when those brands are currently freaked out by the seemingly erratic behavior of the new CEO.

Even if Musk somehow learns to stop live-tweeting his trial-by-fire, the damage might be done: natural iterations on business plans and features will play into a by now near-universal stereotype of Musk as some sort of shoot-from-the-hip madman, which will deter business partners from committing.

(Like, if I'm Fox News or the NY Times, do I want to commit to publish on Musk's new platform when tomorrow he might get stoned and decide to offer hosting to the Daily Stormer? Or, if I'm Bank of America, do I want to commit to consumer payments when tomorrow Musk might pivot his "super app" to offering video shorts instead?)

Again, maybe I'm living in a bubble, but Musk's reputation seems like a huge impediment to, like, doing this. At all. Ever.

So what percentage of budget from, say, the UC system went to administration forty years ago vs today? My point was that this should be public data, yes, surprisingly, is not mentioned.

Per https://nces.ed.gov/programs/digest/d21/tables/dt21_334.30.a..., after two minutes of googling, if I am reading correctly, the percentage spent by four year institutions on “academic support”, which includes admissions, has remained at about 8% from 2000 to 2020. That seems like a basic statistic Silvergate should have on hand, and one that invalidates his hypothesis, right?

I agree they’re not mutually exclusive. But each additional real dollar cost is attributable to one cause or the other. Silver gate seems to only entertain one explanation for all inflation in education, despite very well known alternative hypotheses.

That was my point: not addressing alternative theories, while failing to furnish seemingly readily available supporting evidence, seems on its face dishonest.

There are compelling alternative explanations for the rising cost of education (https://en.m.wikipedia.org/wiki/Baumol%27s_cost_disease#Educ...).

If the underlying cause were primarily administrative overhead, that should be readily apparent in the (public) budget of any public university. I’m not going to go do Silvergate’s research for him, but I’m surprised he doesn’t cite such a readily available piece of compelling evidence, which at best suggests he’s too lazy to track it down, and at worst that reality does not support his hypothesis.

This is, of course, silly trolling. If international law were a tool of US hegemony, then the US would not have been afraid to sign onto the ICC.

The reality is that both America and Russia have willingly violated international law in the past, and accountability—not to mention fairness for smaller independent countries like Ukraine—is hard to come by.

From the perspective of any independent nation in Europe, it’s clear the world is more dangerous than was hoped thanks to Russia’s aggression. You are obviously an American, but not everything is about your country.

Some interesting reading:

https://www.usni.org/magazines/proceedings/2022/october/less... https://www.iiss.org/blogs/military-balance/2022/07/russias-...

tl;dr:

- Low accuracy

- Poor performance (duds/failures)

- Limited inventory