Start Up No.1871: why everything’s a subscription now, art communities ban AI, Peloton’s Wi-Fi trouble, iOS 16!, and more

Obsidian provides some of the sharpest possible edges: useful in the Stone Age, still useful now. CC-licensed photo by jessie essex on Flickr.

You can sign up to receive each day’s Start Up post by email. You’ll need to click a confirmation link, so no spam.

A selection of 10 links for you. Pomp and circumstance. I’m @charlesarthur on Twitter. Observations and links welcome.

The other “MoviePass economy” • The Diff

Byrne Hobart:


Why is it that every company that makes money on transactions seems to be crunching the numbers and concluding that it’s time to launch ThatCompany+, ThatCompany Prime, or ThatCompany One, often at $9.99 per month? Is the future of consumer spending going to be dozens of 5% discounts every month, and a thousand dollars a year of subscription fees to pay for access to them?

The increasing prevalence of subscriptions represents three trends, both of which are tied to the increasing maturity of the tech sector and of tech-enabled retail.

First, as markets get more competitive, the challenge moves from altering customer behavior to winning share among customers who engage in a particular behavior. There was a time when “hail a cab with your phone” was a weird idea, and when “buy it online even if you need it tomorrow” was a ludicrous proposition. Some of these behavioral changes were gradual—the first time I ordered Seamless while working late, instead of just leaving the office to grab some fast food, it felt like the height of decadence. Then it got routine.

Once there are fewer customers to be gained by radically changing habits, there are more customers to be lost from slight shifts in habit, like using a different app to perform the same function. Covid was a massive push towards adopting new consumption habits, especially habits that involved delivery. This has partly receded, but it’s meant that there’s no longer a meaningful population that isn’t at least considering alternatives to brick-and-mortar stores.

Economically, part of what you’re doing when you sign up for an app subscription is that you’re buying your consumer surplus upfront. David Friedman’s Hidden Order has a chapter on exactly this: he imagines a demand curve for cookies such that a price discrimination strategy—buy your first six cookies for $0.70 each and any further cookies for $0.50 apiece—leads to more gross profit per customer.

He then extends this to a more fine-grained price discrimination, with a sliding scale price for each cookie. And then, the magic trick: sell the customer a subscription, price each cookie for subscribers at marginal cost, and the producer captures the entire economic surplus; the supply/demand graph looks like perfect price discrimination applied down to the crumb.


The war between buyer and seller goes on. Explains perfectly why you’re continually being upsold something for just that little bit of money.
unique link to this extract

Online art communities begin banning AI-generated images •

Andy Baio:


On Sunday, popular furry art community Fur Affinity announced that AI-generated art was not allowed because it “lacked artistic merit.” (In July, one AI furry porn generator was uploading one image every 40 seconds before it was banned.) Their new guidelines are very clear:


Content created by artificial intelligence is not allowed on Fur Affinity.

AI and machine learning applications (DALL-E, Craiyon) sample other artists’ work to create content. That content generated can reference hundreds, even thousands of pieces of work from other artists to create derivative images.

Our goal is to support artists and their content. We don’t believe it’s in our community’s best interests to allow AI generated content on the site.


Last year, the 27-year-old art/animation portal Newgrounds banned images made with Artbreeder, a tool for “breeding” GAN-generated art. Late last month, Newgrounds rewrote their guidelines to explicitly disallow images generated by new generation of AI art platforms


The number of AI image generators is starting to grow too fast to keep a handle on. And now people are starting to ban it? That’s going to create a lot of tension.
unique link to this extract

Nikola founder faces securities-fraud trial over promises about electric trucks • WSJ

Corinne Ramey and Ben Foldy on the impending trial over a collapsed EV-promised company:


[Trevor] Milton was an unconventional executive. He said he didn’t finish high school or college but was a serial entrepreneur who started several companies before Nikola. Those ventures often ended up with disputes, litigation and disappointed investors, according to former employees, customers, investors and documents.

…The trial could largely hinge on what Mr. Milton said in television interviews and podcasts and on social media.

On a podcast in 2020, Mr. Milton said that until Nikola came on the market, hydrogen was about $16 a kilogram. “Now Nikola is producing it well below $4 a kilogram,” he said. Prosecutors said that Nikola had never produced hydrogen, at any price.

Mr. Milton also said in interviews that the Badger was a “fully functioning vehicle inside and outside.” When he was asked on Twitter when the first prototype would be produced, he wrote, “Already.” Prosecutors said Nikola had only renderings of vehicles and concept sketches.

In one tweet, Mr. Milton wrote that the Badger would have a drinking fountain using the water created by the truck’s hydrogen fuel cell. Days later, Mr. Milton typed “can you drink water from a fuel cell?” into an internet search, prosecutors alleged.

The rapid climb in Nikola’s shares made Mr. Milton a multibillionaire, based on his holdings. Before stepping down from Nikola, he purchased a $32.5m ranch, the most expensive home in Utah at the time, and a Gulfstream jet.


Among Milton’s other tricks was showing a truck apparently moving under its own steam. (Well, hydrogen. Haha.) In fact it was rolling down an incline.
unique link to this extract

Diffusion Bee is the easiest way to run Stable Diffusion locally on your M1 Mac. Comes with a one-click installer. No dependencies or technical knowledge needed • Github

Divam Gupta:


Diffusion Bee is the easiest way to run Stable Diffusion locally on your M1 Mac. Comes with a one-click installer. No dependencies or technical knowledge needed.

Runs locally on your computer no data is sent to the cloud (other than request to download the weights and checking for software updates).

If you like Diffusion Bee, consider checking , a one-click tool to train machine learning models


Less than two weeks ago, when Stable Diffusion still needed all sorts of packages and command line high jinks, I wondered how soon we’d have a single-app download for MacOS: I thought it would be the end of September. Well, two weeks off. Pile in. Once I’ve had a few moments I’ll see what it generates for The Overspill. Let’s ride the tsunami together!
unique link to this extract

How obsidian Stone Age knives still cut it in surgery • CNN

Peter Shadbolt:


Even today, a small number of surgeons are using an ancient technology to carry out fine incisions that they say heal with minimal scarring.

Dr. Lee Green, professor and chairman of the Department of Family Medicine at the University of Alberta, says he routinely uses obsidian blades. “The biggest advantage with obsidian is that it is the sharpest edge there is, it causes very little trauma to tissue, it heals faster, and more importantly, it heals with less scarring,” he said. “It makes for the best cosmetic outcome.”

He explained that steel scalpels at a microscopic level have a rough cutting edge that tears into tissue, a function of the crystals that make up the metal. Obsidian, meanwhile, cleaves into a fine and continuous edge when properly cut.

Green said he once helped documentary makers produce a program on surgical technology in ancient Egyptian, setting up a blind test on the cutting power of obsidian.

Using cultured-skin burn dressing, a substance composed of skin cells, he made an incision with a modern scalpel and a parallel incision with an obsidian scalpel.

The host of the program was then invited to look at the cuts under a video microscope and tell the difference.

“It wasn’t hard to tell the difference at all. As soon as he turned around, everyone in the studio was like ‘Ohhh,’ ” Green said. “Under the microscope, you could see the obsidian scalpel had divided individual cells in half, and next to it, the steel scalpel incision looked like it had been made by a chainsaw.”


Glass really is remarkable stuff. The sharpest material.. and yet a liquid.
unique link to this extract

New Wi-Fi data shows why Peloton is in trouble • Protocol

Janko Roettgers:


Home workouts are so 2021: as people leave their pandemic cocoons, they’re increasingly ditching their Peloton classes, according to new data from Wi-Fi networking startup Plume Design.

During the first six months of this year, the average amount of data streamed to at-home fitness bikes was down 23% when compared to the same time span a year ago, Plume revealed as part of its IQ Smart Home Market Report released [last] Wednesday.

It’s the single biggest contraction among home Wi-Fi devices that Plume was able to measure. Fitness bikes were followed by media players (Blu-ray players, iPods and similar devices), which saw a decline of 14%, and PCs, which were down 7% year-over-year. The picture looks notably different for modern home entertainment devices. Smart TVs saw data consumption increase by 34% year-over-year, with smart speaker data usage up 27% year-over-year.

The bleak picture for Peloton comes after the company announced plans to lay off 780 employees and close a yet-to-be-determined number of stores earlier this month. Last week, Peloton told investors it had losses of $1.2bn during the most recent quarter, with revenues down 28% year-over-year.

Peloton also saw its member numbers decline by 2% quarter-over-quarter and added close to zero new connected fitness subscribers in its most recent quarter. What’s more, even subscribers who are sticking with the company aren’t working out nearly as much as they used to. The number of average monthly Peloton workouts per subscriber was down 26% year-over-year.

Plume’s data not only corroborates those trends but also shows that connected fitness as a whole is still a pretty small phenomenon when compared to other types of entertainment.


Surprised that smart TVs and speakers saw such a large increase: I’d have expected that it would have held steady compared to last year, or even fallen slightly as people spent more time out of home. Overall, this is only a proxy for activity. (I haven’t looked at the report, so it might contain more wrinkles.)
unique link to this extract

AI read my movie script (and hated it) • Trung Phan

Trung Phan:


I sent Bergquist my really, really funny script and [AI analyser] Corto put it through a multi-step process:

• Ingest and analyze: Corto analyzes and tags script variables including narrative types, emotional tones, character arcs, topics and more. Once the “narrative DNA” of my script is defined, Corto compares it to a database of what Bergquist says is more than 700,000 TV and film titles.

• Generate list of comps: Corto identifies the best matches based on “narrative DNA.” (I was happy to find the classic Southeast Asia comedy “The Hangover II” among the comps.)

• Social media analyses: Corto picks the top 10 closest comps that have grossed at least $50 million and pulls the social media engagement (via Instagram, Twitter, Reddit, TikTok) around these titles. 

• Extract audience segments: Corto examines the commercial potential of my project based on the appeal of the comps to different demographics (age, gender) and which communities to target to help the project go viral. Marvel fans, for instance, are good at getting different communities in their projects; could [Phan’s screenplay] “The Lose” have somehow been marketed to these crowds?

The analysis showed that my film ranked poorly on two scoring systems: uniqueness (how similar was the “narrative DNA” to comps?) and interestingness (did the script have a large character set with a wide range of archetypes?).

The conclusion: Only a star — Corto recommended Chris Pratt — could make my formulaic film successful.


Why stop at Pratt? Go the whole hog – get Tom Cruise to film it. He’s box office platinum. But then he adds:


According to Bergquist, several major film studios are already using the process I just described via Corto’s web-based platform.


Not sure if it’s good or bad that machines are now helping on whether to greenlight stuff.
unique link to this extract

What scientists have learnt from COVID lockdowns • Nature

Dyani Lewis:


Some researchers argue that countries could have avoided blunt all-of-society lockdowns, especially after the measures taken early in 2020. Among them is Mark Woolhouse, an infectious-disease epidemiologist at the University of Edinburgh, UK, who advised the Scottish government during the pandemic. He argues that it might have been possible to avoid the closing of schools and cooping-up of younger people — who were at lower risk of COVID-19 — while focusing efforts on protecting vulnerable and older people as soon as high-risk individuals and settings were identified. “This pandemic was crying out for [a] precision public-health response, because the risks associated with the public-health threat with the virus were so focused on a small minority, and the harms done by things like lockdown were not focused on the same people,” he says.

But many researchers have pushed back against the idea that a more targeted approach was ever possible. Klimek says that roughly one-third of the population in wealthy nations was vulnerable because of underlying health conditions, so targeted measures would have been difficult to implement. And the virus has caused not only deaths but also post-infection illnesses such as long COVID — which has emerged as a health burden even for people who had mild disease.

Another targeted option for governments considering how to reopen societies might have been to keep only high-risk locations closed — restaurants and bars, say, or even neighbourhoods with high population mobility, says Serina Chang at Stanford University in California, who worked with colleagues to identify such places using cellphone data14. But shutting down neighbourhoods would probably disproportionately affect socially disadvantaged communities. “Fairness is such an important question here,” she says.

Woolhouse says there was scant effort to debate the scale of potential harms caused by lockdowns, meaning that policymakers were unable to weigh up costs and benefits properly. Indeed, early on, many countries adopted a ‘save lives at any cost’ approach, he says.


The TL;DR is “it’s complicated, and they don’t really agree, and if something like it happens again they won’t know again.”
unique link to this extract

iOS 16 review: unlocking the lock screen • The Verge

David Pierce:


The lock screen is the true star of iOS 16. Apple has reconceived its purpose altogether, shifting it from just a clock and a bunch of notifications to something much more like a second homescreen. Lock screen widgets were an instant upgrade to my phone life: I can now see my calendar without unlocking my phone or even swiping right to get to that page of widgets everyone always forgets about, and I have a tiny widget that launches a new note in my notes app.

My favorite iOS 16 widget comes from the habit tracking app Streaks. I have “take 5,000 steps” as a daily goal (we’re still in a pandemic, I work from home, and 5,000 steps feels like an accomplishment some days now) and a widget on my lock screen with a meter that slowly fills up as I approach that number. It’s a subtle reminder every single time I look at my phone that I probably need to go outside and touch grass.

The iPhone has never been good at these kinds of light-touch interactions. Before iOS 16, most things required you to pick up your phone, unlock it, swipe to the right homescreen, and open an app. Apple has tried to shrink that process through Siri voice commands, and part of the Apple Watch’s whole appeal is easier access to simple tasks. But “put a bunch of them on your lock screen” might be Apple’s best solution yet. And when you pair it with the always-on displays on the iPhone 14 Pro, the iPhone becomes a fountain of useful information without requiring a single tap.


iPadOS 16 is delayed (because Stage Manager is a mess, it seems). Not sure quite how useful lock screen stuff will be if you don’t have an always-on screen, but certainly better than nothing. It feels like every year we get asymptotically closer to nothing new – Zeno’s OS.
unique link to this extract

Goldman’s Apple Card business has a surprising subprime problem • CNBC

Hugh Son:


Lenders deem bad loans “charge-offs” after a customer misses payments for six months; Goldman’s 2.93% net charge-off rate is double the 1.47% rate at JPMorgan’s card business and higher than Bank of America’s 1.60%, despite being a fraction of those issuers’ size.

Goldman’s losses are also higher than that of Capital One, the largest subprime player among big banks, which had a 2.26% charge-off rate.

“If there’s one thing Goldman is supposed to be good at, its risk management,” said Jason Mikula, a former Goldman employee who now consults for the industry.  “So how do they have charge-off rates comparable to a subprime portfolio?”

The biggest reason is because Goldman’s customers have been with the bank for less than two years on average, according to people with knowledge of the business.

Charge-off rates tend to be highest during the first few years a user has a card; as Goldman’s pool of customers ages and struggling users drop out, those losses should calm down, the people said. The bank leans on third-party data providers to compare metrics with similar cards of the same vintage and is comfortable with its performance, the people said.

Other banks also tend to be more aggressive in seeking to recover debt, which improves competitors’ net charge-off figures, the people said.

But another factor is that Goldman’s biggest credit product, the Apple Card, is aimed at a broad swath of the country, including those with lower credit scores. Early in its rollout, some users were stunned to learn they had been approved for the card despite checkered credit histories.

“Goldman has to play in a broader credit spectrum than other banks, that’s part of the issue,” said a person who once worked at the New York-based bank, who asked for anonymity to speak candidly about his former employer. “They have no direct-to-consumer offering yet, and when you have the Apple Card and the GM card, you are looking at Americana.”


Wonder if some people took advantage to get Apple stuff on the cheap with no intention of paying back.
unique link to this extract

• Why do social networks drive us a little mad?
• Why does angry content seem to dominate what we see?
• How much of a role do algorithms play in affecting what we see and do online?
• What can we do about it?
• Did Facebook have any inkling of what was coming in Myanmar in 2016?

Read Social Warming, my latest book, and find answers – and more.

Errata, corrigenda and ai no corrida: none notified

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.