Stack Overflow on Snowflake Cortex: Answers Without Attitude
For years, developers would go to Stack Overflow to ask and answer technical questions, and the site built up a formidable reputation for having the definitive answers to the almost endless problems that plagued programmers.
That’s until ChatGPT came along in 2022, however, giving devs an even easier way to get answers (many of which, let’s be honest here, were purloined straight from Stack Overflow).
But now that we are firmly in the generative AI (GenAI) era, that huge knowledge base is not going to waste. Cloud data platform provider Snowflake has made available the entire body of work (along with Stack Overflow’s other repositories) so it can be used by its customers in the company’s Cortex AI service.
“This big opportunity for every enterprise is that now they can become an AI app builder, and tap into this vast amount of unstructured content that exists,” noted Prasanna Krishnan, Snowflake’s head of apps and collaboration, in an interview with TNS. “Snowflake was in a really unique position to sort of connect the two sides.”
The dataset will help Cortex users enrich their AI applications and agentic systems. Now, Cortex users can opt to include Stack Overflow’s knowledge base in the sources they use to build out their AI-generated work, through the Snowflake Marketplace.
Answers-As-A-Service
The Stack Overflow website is actually one of over 150 Q&A repositories run by the company, collectively called “Stack Exchange,” that the company also offers to third-party applications to enhance their own chatbots and other knowledge-hungry AI applications (under the name Knowledge Solutions).
The deal with Snowflake means that Stack Overflow data (about 60 million questions worth) is available as an option on the Snowflake Marketplace, as a Cortex Knowledge Extension. The extension allows Snowflake users to incorporate Stack Overflow material into the collection of sources used for their own AI projects and agentic systems.
About half of these questions are from Stack Overflow itself, and the other half comes from the company’s other sites. These additional sites cover other aspects of computing (system administration, security, game development) and nontechnical and technical-adjacent aspects (science and math, arts and recreation, business, home improvement).
“This is the place for the most authoritative content that exists on the internet,” boasted Prashanth Chandrasekar, Stack Overflow’s CEO, in a TNS interview.
A Copilot for Your Business
An enterprise building its own Copilot-like application for developers can include the ability to do a Cortex search call that includes the Stack Overflow extension.
Like a number of companies, Uber ran an internal Stack Overflow site built on Stack Overflow’s enterprise offering, Stack Overflow for Teams, with a chat interface called Uber Genie, which provides answers to questions from within Slack and Uber’s ticketing systems.
The deal with Snowflake gives enterprises an easier way to work with Stack Overflow.
“It’s going direct into the user inside the enterprise, so they don’t have to wait for their company to strike some big data licensing agreement with us,” Chandrasekar said.
Price-wise, Stack Overflow has different tiers of access, depending on what the customer is paying. Using the whole set for data training may come with a bulk price, while another party that may just want to dip into the occasional answer would pay more on a piecemeal basis.
One of the options Snowflake offers is the ability to pay for the data “by the drink,” Krishnan said. This is better for smaller businesses that may not have the money for an upfront enterprise contract.
Stack Overflow’s pricing is based on the token, Chandrasekar added, noting the company has about 50 billion tokens of data to offer.
Snowflake launched its Marketplace in 2019. The company already had the ability to share data across customers via access control. The Marketplace brought the ability for Snowflake customers to pay third parties for the use of the data. Snowflake would go on to extend Marketplace for the sharing of applications, in addition to data.
It was the GenAI applications that made Stack Overflow’s data so much more appealing to Snowflake, because large language models (LLMs) worked well with unstructured data and the Snowflake platform could query the data in natural language, Krishnan said.
The data includes not only the questions and answers from subject matter experts, but also comments, tags and votes. The information is also enriched with metadata. The data can be queried from within Snowflake using natural language.
Snowflake can also provide some additional benefits over going to Stack Overflow directly. For instance, you may not know the exact question you want to ask. In these cases, the intelligence aids with Snowflake Cortex may lead you to data you may not have realized was relevant.
It can also cut down on the snark. Stack Overflow has long been known as sometimes a less-than-friendly place for newbies who ask basic questions that have long been answered before. Commenters are quick to tell users to RTFM. But instead of attitude, Cortex will just put you in the right direction.
A New Business Model
Stack Overflow has been in a tough place since the emergence of GenAI. Web traffic started dropping almost immediately after ChatGPT debuted.
But in 2024, 37.2% of the site’s users checked in daily, and this number had declined to 24.9% in 2025. Only 82% are visiting at least once a month, down from 94% in 2024 (though now the company is seeing an influx of new visitors double-checking work done by AI).
But even as page views dropped, the company noticed another pattern.
“We’ve been the place where people have gone and asked questions.”
— Prashanth Chandrasekar, CEO, Stack Overflow
“We started noticing that a lot of [AI companies] were hitting our APIs really, really hard,” Chandrasekar said.
Soon after ChatGPT came along, all the other major LLM builders scraped Stack Overflow — along with the rest of the internet — for content. “So we had to change the way we approached things,” he said.
So Stack Overflow has been leaning into the partnerships. It also partnered with Moveworks to provide Stack Overflow integration for Moveworks‘ AI assistant. (Moveworks is now being acquired by ServiceNow.) It has also partnered with OpenAI and Google for larger bulk licensing of the data.
Good for the Community, Too
What’s the benefit to contributors for sticking around? It’s a place to build your professional status and recognition, Chandrasekar said.
Stack Overflow plans to increase its investment in its community, giving contributors and commenters a chance to learn more about their craft and receive recognition for the answers they provided through a point system, but also through attribution when their answer is used.
“We want to make sure it’s still open to the community, like an average user or academic — all those people should have free and open access,” he said. “But if you’re a company looking to build something and monetize off of use, you got to pay us so that we can then invest back in the community.”
Update August 11: The post was updated with more information about how the service is monetized.
TNS Research Analyst Lawrence Hecht contributed to this post.
