StackOverflow now gets fewer than 1000 questions a month

Original Article ↗ Hacker News Discussion ↗

Stack Overflow’s monthly question volume has collapsed to fewer than 1,000, prompting debate over whether AI tools or longstanding cultural and design problems bear most responsibility. Many argue that hostile moderation, aggressive duplicate-closing, and gamified reputations had already driven experts and newcomers away years before ChatGPT, while others see large language models as the decisive blow to a Q&A model now easily replaced for most routine programming problems. Commenters also worry about the broader loss of public, searchable technical conversations—whether on Stack Overflow, Reddit, or blogs—and what that means for future human learning and AI training data alike.

Scale and shape of the decline

  • Question volume peaked around 2013–2014, then flattened and declined from ~2017 onward, with a temporary COVID spike in 2020.
  • By 2024–2025, monthly new questions are below the site’s first months, leading some to say SO “effectively no longer exists” as an active Q&A site.
  • Other Stack Exchange sites show similar long-term declines, though at different rates.

AI vs. self‑inflicted causes

  • One camp argues the collapse is “100% because of AI”: LLMs now answer most SO‑style questions instantly, making the product obsolete “like horse‑drawn carriages.”
  • Another camp says SO was already “dying well before AI”: traffic began falling in 2017–2020, before public LLMs, so AI only hastened an ongoing decline.
  • Additional explanations: backlog of “important” questions was mostly solved by mid‑2010s; GitHub issues, docs, and blogs absorbed many project‑specific problems.

Moderation, culture, and gamification

  • Many describe SO as hostile: questions closed as duplicates despite being different or newer, heavy downvoting, nitpicky rule‑lawyering, and power‑tripping volunteer curators.
  • Gamification (rep, badges, review queues) is blamed for incentivizing closing, editing, and rule enforcement over actually helping.
  • Some defenders argue strict curation was essential to keep signal high and that most complaints stem from users ignoring clearly stated rules.

Aging content and technical constraints

  • “No duplicates” policy plus fast‑moving tech led to old, highly‑voted but obsolete answers being canonized while fresh variants were closed.
  • Lack of good support for versions and changing APIs meant answers from 10+ years ago often remained “official,” frustrating users.

Migration to other tools and spaces

  • Many former SO users now default to LLMs plus official docs; some also rely on GitHub issues, Discord/Slack, Reddit, and blogs.
  • Several note that niche Stack Exchange communities (sci‑fi, worldbuilding, DBA, Server Fault) were healthier and more “fun,” but depended on SO’s traffic.

Knowledge commons and AI training

  • Multiple commenters worry that as public Q&A dries up, future models will be trained on “slop” rather than rich human discussion.
  • Others view SO’s content as a foundational dataset that has already been absorbed; the remaining live Q&A is seen as economically and socially undervalued.