Discover more from Social Warming by Charles Arthur
If you tolerate this, then your APIs will be next
Plus the latest hijinks from the world of AI thingamajigs
Perhaps it was all about the airplane tracker? When Elon Musk inducted the porcelain to Twitter, he said that
As a context note under that tweet now notes, “The implied account in this tweet is @elonjet, which is currently banned: [it] used publicly available plane tracking data to follow which airports Musk’s private jet flew to.”
Well, @elonjet is permanently suspended, as of early December. But how vexing for a superbillionaire, that there should exist accounts powered by computers talking to other computers that tell you where people are!
The API—application programming interface, ie the slot in the side of Twitter that lets computer talk directly unto computer, rather than letting human fingers get in the way—seems to be Musk’s least favourite Twitter feature. A couple of weeks ago third-party apps such as Tweetbot and Twitterrific suddenly stopped being able to read or post to the site via the API. For the people who relied on those apps, because they found Twitter’s own one so terrible (🙋♂️), this was quite a blow. A few days later, Twitter tweeted that it was “enforcing its rules” (the apps hadn’t in fact been breaking any then-extant rules) and then retconned its developer agreement to say that yeah, those weren’t allowed now; nobody but Twitter could build a Twitter-like experience using the API.
Which left the question: what’s the API for now, exactly? Cue a certain amount of nervous well-let’s-get-on-with-things-perhaps-the-storm-has-passed whistling on the part of everyone who made use of the API for automated services that post pictures, or tell you when an aircraft has gone past a Raspberry Pi in London, or a cat has gone in (or out) of a catflap, or a zillion other silly and clever and inventive things.
A paid basic tier. Well, that’s going to kill a lot of those silly and clever and inventive things. Everyone at any company that has ever tried to collect money from people knows that there is an absolute chasm between “yup, it’s free” and “it’s very, very cheap, honest”. Even if the charge is 1p per 100,000 API accesses, and your automated account is unlikely ever to roll over to 100,001 tweets and thus take you to 2p of wild untrammelled spending, you still have to hand over your credit card details, and that is the point where people pause, and think “yes, but?”
One of the things that made Twitter so useful was the ability to set up an automated account which could pump out little bits of environmental information, or mark occasions, or generally make the service fun to use. Who needs to know when Big Ben marks the hour? Apparently 400k people are, judging by the follower count of the Big Ben (“entirely unofficial”) account, established October 2009. Will its creator have a good enough reason to pull out a credit card to keep on bonging?
Pay to play
Musk suggested in a tweet on Thursday night (UK time) that the basic price could be $100/month:
Wow. That’s certainly going to kill off a lot of the free automated accounts. As John Gruber observed, that would destroy a lot of what makes Twitter enjoyable:
These accounts freely produce content Twitter users enjoy. If anything, Twitter should be paying them, not the other way around. It’s like if YouTube started charging creators to post their videos.
What the introduction of a “paid basic tier” (and, presumably, “much more expensive paid tiers as well”, given that the present paid tier costs something like $2,500 per month) is not going to do is kill off bots—as in, the annoying automated accounts that, when you use the phrase “hacked account” on Twitter, @ you or DM you suggesting some stupid username on Instagram who “can get your account back”. (They cannot: I once spent an afternoon engaging with the idiots behind this stuff, and they know less about how to get around 2FA than ChatGPT.) It won’t get rid of the pay-for-play influencer networks, or the crypto spammers. It won’t get rid of the state-sponsored (or state-adjacent-sponsored) networks. For a simple reason: the people behind those make money from what they’re doing, so they have a strong financial incentive to keep doing it. In many cases, they’re not going through the API; they’re just running their bots or their searches against a terminal interface. Nor is it hard to write programs that will post into web pages. The only way to stop those is to kill them individually. Such as, say, an account that posts details about where your jet’s going. (Not many people face that problem.)
Unfortunately for Musk, the only way to get close to killing off the troublesome bots, as opposed to the useful bots, is to invest a lot of time, AI and people in identifying and rooting out the fake accounts. It’s tedious! It’s laborious! It’s comparatively expensive both in compute and human resource! But it does make the service a lot nicer to use, in dramatic contrast to actions such as shutting off the API to third-party apps and making all the enjoyable accounts go through existential crises.
The absurd thing is that Musk (advised perhaps by “the goons”, as the venture capitalists and others now populating the decision room were known to the staff) has acted over the API has got the sequence exactly wrong. First he killed the third-party apps; then he rewrote the developer agreement; now he’s introducing a paywall.
sequence of Out
No, no, and no. First you say that you’re going to introduce tiering to the API, and that there will be a limited free tier - say, 10 tweets in a 24-hour period. (Or 12. Or 24. Let Big Ben ring out!) Next up, there will be a paid-for tier, which will be accessible to third-party apps: they’ll have to figure out how to monetise their users to pay for it. Then you have the super-expensive tier, which already exists, that’s used by the companies that do the really hefty analysis of Twitter content.
This gets past a number of problems, such as wanting to have automated accounts that add to the general usefulness of Twitter while not wanting them to push too much content. (Rate-limiting is already a feature of the service; if you set up a new account and start blasting away, your account will get paused very quickly.) It also allows third-party apps, which let’s remember are favoured by the people who get other people to come to the site, to continue to do their work. And by explaining it and introducing it over some reasonable period of time, rather than leaving everyone hanging for a week wondering what the hell is coming next, you keep your developers and users happy.
As it is, a lot of developers who do API-based work have looked at this latest SNAFU and gone even more definitively over to Mastodon. (I’m there too.) There’s a high expectation among developers there that this will kill off a lot of accounts, and I suspect that’s true. (The death knell will be if WeRateDogs—9.3m followers!—goes, and judging by the comments of Matt Nelson, who runs it, in this story, that’s getting more and more likely.)
I feel as though I’m writing the same piece again and again: that Twitter is giving us a strange insight into a platform that is trying its very hardest to drive social warming, by thrusting the For You tab of algorithmic rage in peoples’ faces, by forcing them to use a semi-broken app, by not showing them all their mentions (through incompetence rather than design); yet at the same time it’s dismantling the very things that made it a successful platform, and building walls higher and higher around itself each day. In business, sequencing counts, and the debacle over the API suggests to me once more that this isn’t going to end with a cornucopia of money.
Glimpses of the AI tsunami
(Of the what? Read here.)
• ChatGPT hits 100 million accounts—the fastest growing user base ever. (Previous recordholder TikTok took nine months.) Plus there’s also going to be a $20/month subscription service with faster response and “priority access to new features”. Stand by for the AI-soaked web.
• Why are AI-generated hands so messed up? Yeah, why? Because hands tend to be hidden, so the systems have to guess at those flappy things that radiate off the arms.
• AI-generated voice firm clamps down after 4chan makes celebrity voices for abuse. I mean, that pretty much sums it up? Not the smartest move by the firm involved to make the voice generator an open beta.
• Infobot “provides weekly AI-generated newsletters written specially for you, your investing thesis, and your portfolio. We read and synthesize 700,000 words per week so you don't have to.” Pricey, and I’m not that sure that the newsletters (example) are really that great.
• Buzzfeed’s CEO Jonah Peretti explains how AI-generated content is going to be an integral part of, uh, Buzzfeed in the future: “If the past 15 years of the internet have been defined by algorithmic feeds that curate and recommend content, the next 15 years will be defined by AI and data helping create, personalize, and animate the content itself.” Hey, that’s sort of what I said!
• Google has an AI music maker: generate music from text (eg “a calming violin melody backed by a distorted guitar riff”). I feel like we’ve been doing this quite a lot for years, but this is apparently a good new one. (It takes too long to load for me. Fast connections only.)
You could also sign up for The Overspill, a daily list of links with short extracts and brief commentary on things I find interesting in tech, science, medicine, politics and any other topic that takes my fancy.
• Back next week! Or leave a comment here, or in the Substack chat.
You reached the very end! Is the box ticked?