Decoder with Nilay Patel: Stack Overflow users don't trust AI. They're using it anyway

Date: December 15, 2025
Guest: Prashanth Chandrasekhar, CEO of Stack Overflow
Host: Nilay Patel, The Verge Editor-in-Chief

Episode Overview

This episode explores the post-ChatGPT world for Stack Overflow—a platform historically at the center of software development communities. Nilay Patel discusses with CEO Prashanth Chandrasekhar how generative AI has upended Stack Overflow’s business, community, and technology, and how the company is shifting its focus toward enterprise SaaS and data licensing. The episode underscores the major tension: AI is everywhere, millions use it, but very few trust it.

Key Discussion Points

1. The “Code Red” Moment: AI’s Disruption of Stack Overflow

• ChatGPT’s launch in 2022 triggered an existential crisis for Stack Overflow, destabilizing the core Q&A community.
• Prashanth’s immediate response: Carved out 10% of company resources (~40 people) for a dedicated team to rapidly address generative AI’s impact, setting a six-month deadline for an initial response.
  • “We went into what is the equivalent of a Code Red situation inside the company. It was an existential moment, especially for our public platform.” —Prashanth (05:59)
• Source of the “crisis mode” idea: Prior experience at Rackspace responding to AWS, and lessons from Clayton Christensen’s “Innovator’s Dilemma.”
• Communication during this period was a series of transparent memos, not just one dramatic announcement.
• The company shifted focus toward two main businesses: the public Q&A platform and an enterprise SaaS product (private Stack Overflow for businesses).

2. AI’s Double-Edged Impact: Inputs & Outputs

• Inputs: AI-generated, often low-quality (“AI slop”) answers flooded the site. Community and moderators quickly detected and flagged ChatGPT-style answers, prompting a ban on all AI-generated answers.
  • “Our proposition is to be the trusted, vital source for technologists. There are only a few places where you can go where you're not dealing with AI slop.” —Prashanth (15:33)
• Outputs: More developers simply use AI models directly instead of searching for human answers on Stack Overflow.
• Stack Overflow’s response:
  • Maintains a strict ban on posting AI-generated content.
  • Launches features like AI Assist—a conversational AI interface grounded on Stack Overflow’s 90M Q&As but designed to keep human curation central.
  • Opens up new types of discussion and community interaction, e.g., subjective questions, group chats, and coding challenges.

3. Reinventing Stack Overflow’s Business Model

• Three-pronged monetization:
  1. Enterprise SaaS (Stack Overflow Internal): Adopted by ~25,000 companies; serves as a trusted knowledge base powering internal AI assistants (e.g., Uber Genie).
  2. Data Licensing: AI labs (OpenAI, Google, Databricks, Snowflake, and more) now pay recurring fees for Stack Overflow data used in LLM pre-training, post-training, RAG, and other applications.
     • “Now we've struck effectively partnership agreements with every single AI lab that you can think of for the most part.” —Prashanth (19:06)
  3. Advertising: Now only about 20% of revenue, targeting major tech companies.
• Stack Overflow is becoming “headless”—integrating wherever developers work (IDEs, code tools) and acting more as an API or knowledge backend.

4. Community Backlash & The Trust Paradox

• Stack Overflow’s core community (“the 1%”) strongly opposes AI-generated answers and data deals with AI companies (e.g., OpenAI).
  • Some users deleted their posts in protest before data licensing; moderators revolted over inability to act fast on “AI slop.”
• Despite this backlash, over 80% of users already use or want to use AI for coding (31:36, 58:13).
• But only ~29% actually trust AI’s output.
  • “80 plus percent of our user base wants to use AI or already using AI for code related topics, but only 29% of that population trusts AI.” —Prashanth (58:13)
• This enormous gap—high usage but low trust—is mirrored across tech communities and is “unlike any other split” Nilay has seen in tech.
  • “Everyone says they don't like it and then we see the numbers and everyone's kind of using it anyway.” —Nilay (57:12)
• Prashanth sees this as resulting from skepticism toward AI’s accuracy, fear of job displacement, and the probabilistic (not deterministic) nature of LLMs.

5. The Future of the Platform and Software Development

• Stack Overflow is expanding community features to provide value beyond Q&A—e.g., hackathon “challenges”, job partnerships (with Indeed), real-time chat with experts.
  • “We want to give them more reasons to come to the site [...] a very popular feature on the site where people are engaging with other experts.” —Prashanth (27:58)
• Prashanth envisions Stack Overflow as a “trust layer” for both public internet and enterprise models.
• He predicts 2026 will be the “year of rationalization”—moving from AI hype to enterprise demands for demonstrable ROI from AI investments, likely leading to consolidation among AI vendors.

Notable Quotes & Segments

• On the existential AI moment for Stack Overflow
  • "[ChatGPT] was an existential moment, especially for our public platform...here you go, you have this really, really slick interface, that natural language interface that allows you to [get answers] on a moment's notice." —Prashanth (05:59)
• On the community’s resistance to AI
  • "There will always be a vocal minority that will push back against incorporation [of AI] because there's a lot of [...] concern about what all this does to jobs and if we let the, you know, cat out of the bag, then what's going to happen.” —Prashanth (34:57)
• On data licensing and the changing internet
  • "The model of the Internet has literally been turned upside down. People relied on [...] search engines and websites and you monetize off of ads. That is completely [changed]...so what we have to do is adopt a new business model to survive and thrive." —Prashanth (36:27)
• On the trust gap
  • "80 plus percent of our user base wants to use AI or already using AI for code related topics, but only 29% [actually] trust AI... And trust is a very deep word." —Prashanth (58:13)
  • "I think developer audiences are very discerning, analytical, and can be prickly if things are not...deterministic, the way it has been for a very long time." —Prashanth (58:13)
• On Stack Overflow’s new strategy
  • "We are building this enterprise knowledge intelligence layer for companies to truly use AI agents in a trustworthy way." —Prashanth (70:29)

Timestamps for Important Segments

| Timestamp | Topic/Quote | |-----------|-------------| | 05:59 | Prashanth describes the “Code Red” moment after ChatGPT launches | | 08:14 | How the “Code Red” was communicated and organized internally | | 14:32 | Impact of AI on Stack Overflow’s input/output—banning AI answers and fighting “AI slop” | | 19:06 | Business pivots: SaaS, data licensing, and ad revenues | | 27:23 | Will new generations of coders use Stack Overflow at all? | | 31:36 | Backlash from community moderators/users and the 80%/29% usage vs. trust data | | 34:57 | Managing the tension between community ideals and business evolution | | 36:27 | On the necessity of data licensing and changes in the internet’s business model | | 41:53 | Negotiating data licensing with AI companies like OpenAI | | 44:48 | Structure and logic of licensing recurring revenue from AI labs | | 54:44 | Stack Overflow’s headcount and financial state in 2025 | | 57:12 | Nilay on the unique usage vs. trust split with AI (“unlike any other split” in tech) | | 58:13 | Prashanth explains developers’ skepticism—but widespread adoption—of AI | | 60:42 | Launch of Stack Overflow’s AI Assist and addressing trust in AI’s output | | 62:37 | Will LLMs keep getting better? Prashanth bets “the compounding effects are very real.” | | 68:02 | Predicting the “age of rationalization” (AI bubble correction) in 2026 | | 70:29 | Stack Overflow’s future focus: enterprise AI agent trust layers and public community |

Final Takeaways

• Stack Overflow is actively transforming—from a public Q&A hub to an enterprise knowledge layer and data licensor for the AI industry.
• Despite user resistance and distrust, AI is thoroughly embedded in how developers work, and the tension between community ethos and commercial imperatives is as sharp as ever.
• Stack Overflow’s future hinges on maintaining trust, both as a community and as a data provider to the AI ecosystem, while navigating a rapidly rationalizing AI market.

End of Summary