In the spring of 2023, a developer named Marco — someone I'd traded emails with for years about the finer points of rate limits and OAuth flows — sent me a message that read like a eulogy. His app, a quiet little tool that helped researchers archive public conversations around climate policy, had just stopped working. Not because of a bug. Because Twitter had pulled the free tier of its API with roughly 48 hours of practical notice, and the replacement pricing started at $100 a month for a level of access that, six months earlier, had cost nothing. "I'm not angry," he wrote. "I'm just done."
I kept that message open in a tab for three days before I knew what to do with it.
The Infrastructure That Developers Built Their Lives Around
To understand what happened, you have to go back further than Elon Musk's acquisition in October 2022. The Twitter API, in its earliest public form, appeared around 2006 — almost as an afterthought, a side door left open by engineers who wanted to see what outside developers would build. What they built was, in retrospect, extraordinary. Third-party clients like Tweetie and Twitterrific gave the platform much of its early design language. Researchers at universities built datasets from the firehose that informed a decade of social science. Journalists wrote bots that tracked congressional votes, wildfire perimeters, earthquake alerts. The platform's cultural weight — the thing that made Twitter feel essential — was partly constructed by people Twitter never paid.
This is the part of the retrospective analysis of Twitter API changes that tends to get lost in the noise of the acquisition drama: the dependency ran both ways. Twitter needed the ecosystem as much as the ecosystem needed Twitter. Developers weren't just users. They were, in a very real sense, unpaid infrastructure workers who had accepted platform risk as the cost of admission.
For years, that risk felt manageable. The API had its frustrations — rate limits tightened in 2012, the whitelist system for elevated access was notoriously slow, the v1.1 transition in 2013 killed several beloved clients — but the underlying covenant held. Free access to public data, in exchange for building things that made Twitter more useful and more visible. The company enforced this covenant imperfectly and sometimes capriciously, but it enforced it.
When the Price of Trust Became Literal
The first real tremor came in 2018, when Twitter began restricting what third-party clients could do with the streaming API, effectively kneecapping apps like Tweetbot from offering real-time timelines. The stated reason involved spam and automation abuse — legitimate concerns, genuinely — but the timing, arriving just as Twitter was pushing its own app harder, struck developers as convenient. The trust that had accumulated over a decade developed its first visible crack.
Then came the acquisition, and the pace of change became something else entirely. Within months, the free API tier was gone. The basic tier, at $100 a month, offered a fraction of the previous free access. The enterprise tier — the one that gave researchers and journalists anything approaching the old firehose — ran into tens of thousands of dollars annually. Companies that had built entire products on the assumption of continued access found themselves in an impossible negotiation: pay rates that made their business models nonsensical, or walk away from years of work.
Many walked away.
What followed was a kind of dispersal. Researchers migrated toward Mastodon's API, which remains genuinely open, or toward Bluesky's AT Protocol, which was designed from the start with federation and data portability as first principles rather than afterthoughts. Some simply stopped studying social media dynamics altogether, which is a loss that won't show up in any earnings call but will show up, eventually, in our collective understanding of how information moves through networked publics.
The Argument Twitter's New Owners Would Make
It would be unfair — and bad analysis — to treat the pricing changes as purely destructive without acknowledging the argument on the other side. Twitter had, for most of its existence, been a company that struggled to monetize at the scale its cultural influence suggested it should. The API was genuinely being abused: scraping operations, spam networks, and data brokers were extracting enormous value from the firehose while contributing nothing. The free tier, in practice, had become a subsidy for commercial actors who had no intention of building anything that benefited the platform.
Putting a price on access is not, in itself, wrong. Charging for commercial use while preserving academic and journalistic access is a model that other platforms have implemented with reasonable success. The problem wasn't the principle. The problem was the execution — the speed, the lack of granularity, the absence of any meaningful carve-out for non-commercial research, and the communication style that veered between silence and contempt.
A retrospective analysis of Twitter API changes has to hold both things at once: the legitimate business case for monetizing data access, and the catastrophic mishandling of a transition that burned through decades of goodwill in a matter of weeks.
What the Rubble Tells Us
Sitting with all of this now, what strikes me most is not the drama of the moment but the structural lesson underneath it. Platforms are not neutral utilities. They are governed spaces, and the governance terms can change at any time, for any reason, with whatever notice the owner considers sufficient. Developers who build on platforms are, in a legal and practical sense, tenants — and the lease has always included a clause that lets the landlord renovate without warning.
We knew this, abstractly. The Twitter API changes made it visceral.
The deeper question — the one Marco's message kept nudging me toward — is what it means to build on someone else's infrastructure when that infrastructure has become genuinely important to public life. Twitter's API wasn't just a developer convenience. It was part of the machinery through which journalists tracked disinformation, through which epidemiologists monitored sentiment during health crises, through which ordinary people built tools that helped other ordinary people navigate a confusing world. When that access disappeared, it didn't just inconvenience developers. It removed capabilities that had become, quietly and without anyone quite deciding it, part of how we collectively understand ourselves.
That's not a problem any single company can be asked to solve on behalf of society. But it is a problem society probably needs to start taking seriously — the question of what happens when critical informational infrastructure is owned by private actors who have no obligation to maintain access and no incentive to consider the downstream effects of closing the door.
Marco eventually rebuilt a smaller version of his tool using a combination of scraped public data and the Mastodon API. It doesn't do everything the old version did. The dataset is thinner, the coverage patchier. He told me recently that he thinks of it as a "good enough" solution, which is how a lot of people describe the things they build after a platform has taken something away from them.
Good enough. That's what the retrospective analysis of Twitter API changes ultimately leaves us with — not a clean lesson, not a villain and a hero, but a quieter reckoning with the cost of building trust on borrowed ground. The question worth sitting with isn't whether Twitter made the wrong call. It's whether we, as developers and researchers and builders, were ever asking the right questions about what we were building on top of.