Stack Overflow is dying: questions fall to 2008 levels
Questions on Stack Overflow have dropped from 200,000 to 1,400 a month: AI is devouring the very source that feeds it, opening a degradation loop.
English · Original discussion in Spanish · Published
Stack Overflow is running out of questions: from 200,000 to 1,400 a month
Stack Overflow is dead. Or at least that's what the numbers circulating since July 2026 suggest. New questions have fallen from around 200,000 a month to barely 1,400, a level not seen since 2008. The cause seems clear: programmers no longer ask the platform, they ask AI.
But the diagnosis needs nuance. The decline started earlier. The platform was already losing traction before the rise of AI assistants; there was only a small uptick in the year of ChatGPT. Part of the queries were repetitive and solvable with the documentation. AI didn't invent the trend, it turned it into a catastrophe.
The paradox: AI feeds on the web it is killing
Large language models have been trained, to a large extent, on answers from Stack Overflow and public repositories like GitHub. Some estimates put the contribution of those sources at 90% of the answers assistants give today.
The problem is twofold. If humans stop asking and answering, no new knowledge is generated. AIs don't create, they combine; they cannot document a bug no one has found yet. When new frameworks and updates appear, there will be no fresh answers to crawl. The models will repeat outdated or outright incorrect solutions. It's the degradation loop that several analyses already anticipate.
A $1.8 billion asset deflating
Prosus paid $1.8 billion for Stack Overflow on June 2, 2021. The context was different: the web was the world's reference for solving coding doubts. Today's comparison is devastating. A website with 700,000 monthly unique visits was able to generate 333 questions a day; Stack Overflow now hardly reaches that figure in several weeks.
The historical archive, with millions of answers, has been exploited in a way that never compensated those who created it. The question of copyright over those answers, used to train models that compete with the platform itself, remains unresolved. Without compensation, the incentive to explain problems publicly fades away. Open knowledge risks becoming a private asset of a few corporations.
What if AI runs out of food?
If the flow is cut off, the next models will be trained on a corpus frozen in time. The ability to answer questions about new technologies will degrade, and it's unclear whether big companies will want to keep knowledge open. Stack Overflow may never be what it was. The most uncomfortable question is whether AI will move forward when its main source of food is dead.
Stack Overflow is dead. Or at least that's what the numbers circulating since July 2026 suggest. New questions have fallen from around 200,000 a month to barely 1,400, a level not seen since 2008. The cause seems clear: programmers no longer ask the platform, they ask AI.
But the diagnosis needs nuance. The decline started earlier. The platform was already losing traction before the rise of AI assistants; there was only a small uptick in the year of ChatGPT. Part of the queries were repetitive and solvable with the documentation. AI didn't invent the trend, it turned it into a catastrophe.
The paradox: AI feeds on the web it is killing
Large language models have been trained, to a large extent, on answers from Stack Overflow and public repositories like GitHub. Some estimates put the contribution of those sources at 90% of the answers assistants give today.
The problem is twofold. If humans stop asking and answering, no new knowledge is generated. AIs don't create, they combine; they cannot document a bug no one has found yet. When new frameworks and updates appear, there will be no fresh answers to crawl. The models will repeat outdated or outright incorrect solutions. It's the degradation loop that several analyses already anticipate.
A $1.8 billion asset deflating
Prosus paid $1.8 billion for Stack Overflow on June 2, 2021. The context was different: the web was the world's reference for solving coding doubts. Today's comparison is devastating. A website with 700,000 monthly unique visits was able to generate 333 questions a day; Stack Overflow now hardly reaches that figure in several weeks.
The historical archive, with millions of answers, has been exploited in a way that never compensated those who created it. The question of copyright over those answers, used to train models that compete with the platform itself, remains unresolved. Without compensation, the incentive to explain problems publicly fades away. Open knowledge risks becoming a private asset of a few corporations.
What if AI runs out of food?
If the flow is cut off, the next models will be trained on a corpus frozen in time. The ability to answer questions about new technologies will degrade, and it's unclear whether big companies will want to keep knowledge open. Stack Overflow may never be what it was. The most uncomfortable question is whether AI will move forward when its main source of food is dead.
Also available in: Deutsch
Summary of a discussion on Burbuja.info - Foro de economía, actualidad y política., translated from Spanish and reviewed before publication. Read the full discussion (34 replies).
More summaries
-
Cheap AI and the End of Illustrative Art
AI enables complex comic creation via simple commands, sparking debates on art devaluation and copyright issues.
-
ChatGPT can't tell a fact from a casual opinion
Generative AI predicts words, it doesn't think: two years of promises about replacing half the economy, and it still confuses a fact with a barstool opinion
-
Walmart retires employee with flowers after 44 years
Gayle retired from Walmart after 44 years with a flower bouquet. The gesture reignites debate on workplace recognition and pension plans.
-
Unverified Claim Links PSOE to Mosque Voter Mobilization in Spain
A viral post claims the PSOE coordinated with mosques for Spanish-Moroccan voters. No evidence exists, and the cited election date is invalid.
-
Sex displaces shared laughter as dominant social value
Shared laughter fades while sex goes viral online. An economic and psychological review of desire sublimation and the cost of instant gratification.
-
Pro-Israel Lobby Loses Image, Not Power, in US Elections
Records show 247 Democratic candidates received $133.9 million from AIPAC and pro-Israel PACs. The lobby’s image has shifted, but its influence remains intact.