Rise and fall of Stack Overflow
- YT :: https://www.youtube.com/watch?v=Bxv1_Nk1k5o
- Original title :: The Death of Stackoverlow | TheStandup
A Standup clip (Prime, Casey Muratori, Trash, Teej) on the history and current state of Stack Overflow: from 207,000 new questions a month at its 2014 peak to fewer than 1,000 in September 2026. The hosts mix the timeline - acquisition, ChatGPT, AI data-licensing deals, the private-equity write-down - with personal stories about chasing reputation and getting a question deleted. The conclusion: as a community Q&A site it is a ghost town, but as a business it is finally (barely) profitable by selling data and private Q&A to enterprises. Full episode: https://youtu.be/cZ-05LxT8Xc.
Timeline
- 2008 :: launched by Jeff Atwood and Joel Spolsky as a free Q&A site for programmers, with reputation earned through up/down votes. Grew into the Stack Exchange network: 183 sites (Super User, math, Vi, Emacs, even Tolkien lore), 24.2 million questions.
- March 2014 :: peak engagement, 207,000 new questions in a single month - all typed by humans.
- 2018 :: Teams enterprise product.
- 2021 :: acquired by Prosus (private equity, Europe) for $1.8 billion - before the AI era.
- 2022 :: Jobs and Developer Story sections shut down.
- 2023 :: ChatGPT makes asking a human pointless; Stack Overflow tries to ban AI-generated answers, moderators strike, layoffs follow.
- 2024 :: the data is sold for AI training - Google API deal in February, OpenAI deal soon after. Users who deleted their answers in protest got suspended; content is CC BY-SA and the site owns it.
- Now :: new questions down 99.5% from peak (5,000 in September 2025, 966 in September 2026). Prosus wrote down $1.5 billion (87%); Stack Overflow is valued at ~$230 million. Fiscal 2026: $129 million revenue and its first ever operating profit, about $1 million - from licensing and enterprise private Q&A.
Personal stories
- Prime :: early user (2009-2010, PHP on XAMPP in NetBeans); 4,700 reputation, mostly from "Pythonic" questions.
- Favourite answer :: the famous Vim answer about not memorising a thousand leaf-node shortcuts, but learning the underlying composable language. It changed how Prime approaches any tool: find the small building blocks that compose into the many cases.
- Trash :: around 2012 spent a year or two farming reputation in niche database communities - answering, even posting and self-answering niche questions - treating reputation as job leverage and networking. Reached about 800 and never actually used it to get a job.
- Casey :: used it only via search results; esoteric questions (GCC switches, undocumented Microsoft behaviour) often had the question asked but no answer. Asking felt intimidating because of the high bar against duplicates.
- Deleted question :: Prime's "best way to measure your dict in bytes" (about a real LRU memory problem from the Falcor days) was deleted by a moderator as deliberately crafted double entendres - with a human-written note complimenting how thoroughly they were woven in.
Emacs vs Vim
The Stack Exchange numbers: Emacs has 38,000 users and 28,000 questions, Vi has ~53,000. Prime claims Vim is more popular and less confusing; Casey counters that Emacs questions cover things Vi users would never expect from a text editor ("M-x toast-mode"), and that a typical Emacs user knows about 5% of it.
Today
Browsing newest questions shows a trickle of niche ones (nftables, paginated SKU listings) drawing downvotes. The hosts see a possible future in internal company Q&A where agents ask and answer each other, though agents can now just read the repos directly. Verdict: the reports of its death are exaggerated as far as money goes, but the community is cooked.