I was three tabs deep trying to find out if anyone had posted photos from a meetup I couldn't make it to yesterday when Twitter just... stopped. Not down-down, no fail whale (RIP), just a wall of text telling me "Rate limit exceeded." I'd apparently read my allotment of tweets for the day, whatever that meant, and the app wanted me to wait it out.
Turns out this wasn't a bug. Elon announced it himself Saturday afternoon: unverified accounts capped at 600 tweets read per day, verified accounts at 6,000, and brand new unverified accounts throttled down to 300. The stated reason was "extreme levels of data scraping" — basically, AI companies hoovering up tweets to train language models, and Twitter trying to make that more expensive to pull off at scale. Within a few hours the numbers got bumped up (800/8,000/400, if I'm remembering the second round right), which tells you roughly how much thought went into the original figures.
I don't even use Twitter that heavily anymore, maybe twice a day, mostly to see if anything's on fire in my little corner of tech, and I still hit the wall. A friend who actually doomscrolls for a living got locked out before 11am.
The funniest part, and I say this as someone who's watched a lot of platform meltdowns from the sidelines since I started this blog back in 2011, is that the stated reason and the actual effect don't point the same direction. If you wanted to stop bots and scrapers you'd throttle by behavior pattern, not slap a blanket daily cap on every logged-in account, including the guy who's been on the platform since 2009 and checks it from his phone at red lights.
What's more interesting to me than the outage itself is what it's nudging people toward. I saw more Mastodon instance links in my feeds this weekend than in the past six months combined, and there's been a steady drumbeat of reporting lately about Meta building its own text-based app to go up against Twitter directly, tied into Instagram logins, reportedly under an internal codename. Nothing's live yet as far as I can tell. But the timing of Twitter setting itself on fire right as a very well-funded competitor is apparently getting ready to ship is almost too convenient to be an accident.
I'll admit I have a soft spot for the old RSS-reader instinct here. This blog predates my Twitter account in spirit if not in actual days, I was already the kind of person who'd rather subscribe to a dozen blogs and read them on my own schedule than sit in somebody's algorithmic feed waiting to be told what mattered today. Getting rate-limited for an afternoon didn't disrupt my life much. It just sent me back to my actual RSS reader (still Feedly, still not thrilled about it since Google Reader died, but here we are), and I got just as much useful information without anyone telling me I'd read too much of it.
None of this is a crisis. Nobody's timeline breaking for a weekend is going to be the thing that finally kills Twitter, people have been predicting that since roughly 2009 and it's still here. But it's a genuinely odd thing to watch a company respond to a stated technical problem with a fix that mostly just annoys its actual users, paying and non-paying alike, in more or less equal measure. If the real goal was slowing down AI training pipelines, I'd bet the big scraping operations had already cached whatever they needed weeks before Saturday, and this weekend mostly just cost Twitter some goodwill from people like my friend who hit her limit before lunch.
I don't have a tidy answer for where this goes next. I'm just going to keep an eye on my tweet count for the rest of the weekend, apparently, like it's a data plan from 2009.