APIs Used to be Free and Open. Then, they got expensive.
Medium deprecated their API.
Twitter locked theirs behind a paywall.
Reddit introduced a pre-approval process.
Then, there’s Stack Overflow. LinkedIn. And so on.
These can’t just be isolated incidents.
The timing is too consistent. The direction too uniform. We are watching a real-time exodus where platforms close their doors to open data policies, each citing its own rationale: quality control, infrastructure costs, abuse prevention, developer ecosystem reform. Whilst, some do it quietly.
The pattern is clear. The free, open data world developers once built inside has morphed into something far more restricted, far more monetised.
When Building and Growing a Community Used to Mean Something
To understand why this ended, go back to the late 2000s through the early 2020s.
The prevailing logic in tech was simple: Grow First, Monetise Later.
Build. Adopt. Experiment. Don’t worry about revenue yet. Burn cash to grow the community. Get more users. Get developers to adopt your stack. Accumulate data. Revenue was a conversation for after you had monopoly on the market.
Underneath the goodwill and the genuine excitement of building things, open APIs were implemented as a growth strategy.
When Twitter opened its API in 2006, third-party developers built clients, analytics tools, scheduling software, and integrations that Twitter itself had no capacity or intention to build. Those tools brought users onto the platform, kept them there, and created an ecosystem that made Twitter feel indispensable. Developers were doing Twitter’s product work for free, motivated by the excitement of building on a fast-growing platform and the reputation that came with being an early mover.
Reddit’s public JSON endpoints and Facebook’s early Graph API operated on the same logic. Medium launched their own API in 2015 on top of its publishing infrastructure for free. The goal? Adoption. Community first, business later.
This was the golden era. It felt like genuine openness, and in a way it was. The ideology and the business incentive pointed in the same direction. Data was free because there was no immediate cost to making it free.
Then the Cost of Capital Went Up
By 2022, that changed.
Inflation drove the shift. Pent-up demand, supply chain disruptions from the lockdowns, and years of loose monetary policy had driven asset prices to unsustainable levels. Central banks raised interest rates at a pace not seen in decades. The era of cheap money ended.
For tech companies, this was existential. The growth-at-all-costs model depended on investors willing to fund losses indefinitely in exchange for future upside. When the cost of capital rose, that patience evaporated. Companies that had been burning cash on developer ecosystems and free API access suddenly needed a credible path to profit.
Every asset on the balance sheet got scrutinised. And data — the accumulated content of millions of users posting, commenting, writing, and arguing on these platforms for a decade — turned out to be worth something.
The API that was once a free growth tool became a revenue line item. Or a cost to be eliminated.
Then AI Made the Data Worth Fighting Over
The rate rises forced businesses to prioritise profit. AI going mainstream in 2023 accelerated that process and changed the stakes entirely.
Large language models need training data. Specifically, they need vast quantities of high-quality human-generated text. And they got it. All thanks to the earlier efforts of these platforms building up communities and encouraging everyone and everybody to share their knowledge, expertise and information. From Reddit threads, Medium articles, Twitter conversations to Stack Overflow answers. This was exactly the corpus AI companies needed, and platforms had been hosting it for free for over a decade.
The implication landed fast. A platform’s historical content was no longer just content. It was a training asset. And if an AI company could ingest it, they could build a product that answered the questions users came to the platform to ask, without those users ever needing to visit other external sources.
That is an existential threat. Not a nuisance. Not a compliance issue. An actual threat to the reason the platform exists.
This is the final act that pushed platforms to close their doors. If AI companies could extract value from that data, then the data had a price. And access should cost something proportional to that value.
Enshittification Was Always the Destination
Cory Doctorow named the pattern in late 2022. The timing of this felt perfect.
Platforms follow a predictable arc. They launch by being genuinely good for users. They acquire and monopolise. Once they have sufficient scale, they shift to serving the businesses that pay to reach those users. Then they begin extracting value from everyone while giving back as little as possible.
First, they serve users. Then, they serve the businesses that pay to reach users. And finally, they serve themselves.
What Survives Is Structural, Not Ideological
Some platforms remain open. Wikipedia. Hacker News. Government data portals. A handful of others. What they share is not a commitment to openness as a value. What they share is a structural reason why closing would cost them more than staying open.
Wikipedia’s legitimacy depends on being universally accessible. Restricting access to its data would undermine the mission that keeps volunteers contributing. The openness is load-bearing.
Hacker News is a community tool run by a company that does not need to monetise it directly. The API exists because Y Combinator’s business is deal flow and reputation among developers. Keeping developers happy is the product.
Reddit’s legacy JSON endpoints survive not by ideology but by inertia. Removing them would break search engine indexing, embeds, and internal tooling the platform still depends on. They persist because the cost of removal exceeds the benefit. That is a different kind of open. Fragile, undocumented, and subject to change the moment the calculus shifts.
The lesson is not that open platforms are dead. It is that openness without a structural reason to exist is temporary.
The Irony Worth Sitting With
The thing that accelerated the API lockdowns — AI — is itself operating on the same logic that created the open era in the first place.
Billions in investment capital. Losses at scale. A promise of future monetisation that has not fully materialised. The same bet that funded a decade of free APIs is now funding the technology that made those APIs worth locking.
That is not a coincidence. It is the same pattern. And there are strong reasons to believe AI is heading down the same path.
And that path is enshittification.