Stack Overflow Users Don’t Trust Ai. They’re Using It Anyway

Today, I’m talking pinch Prashanth Chandrasekar, who is nan CEO of Stack Overflow. I past had Prashanth on nan show successful 2022, 1 period earlier ChatGPT launched. While nan generative AI roar had tons of effect connected each sorts of companies, it instantly upended everything astir Stack Overflow successful an existential way.

Stack Overflow, if you’re not acquainted pinch it, is the mobility and reply forum for developers penning code. Before nan AI explosion, it was a thriving organization wherever developers asked for and received thief pinch analyzable programming problems. But if there’s 1 point AI is bully at, it’s helping developers constitute codification — and not conscionable constitute code, but create full moving apps. On apical of that, Stack Overflow’s forums themselves became flooded pinch AI-generated answers, bringing down nan value of nan organization arsenic a whole.

You’ll perceive Prashanth explicate that it was clear much aliases little from nan jump really large a woody ChatGPT was going to be, and his consequence was axenic Decoder bait. He called a institution emergency, reallocated astir 10 percent of nan unit to fig retired solutions to nan ChatGPT problem, and made immoderate beautiful immense decisions astir building and statement to navigate that change.

Three years later, Prashanth says Stack Overflow is now very comfortable chiefly arsenic an endeavor SaaS business, which provides AI-based solutions that are tailored to different companies’ soul systems. Stack Overflow besides operates a large information licensing business, trading information from its organization backmost to each those AI companies, ample and small.

That’s a beautiful large pivot from being seen arsenic nan spot wherever everyone tin spell to get thief pinch their code. So I had to inquire him: does Stack Overflow moreover pull caller users anymore, successful 2025, erstwhile ChatGPT tin do it each for you? Prashanth said yes, of course. You’ll perceive him explicate that while AI tin grip elemental problems, for thorny, analyzable ones, you really want to talk to a existent person. That’s wherever Stack Overflow still brings group together. 

You’ll perceive america travel backmost to a single stat successful particular: Prashanth says much than 80 percent of Stack Overflow users want to usage AI aliases are already utilizing AI for code-related topics, but only 29 percent of that organization really trusts AI to do useful work. 

That’s a immense split, and it’s 1 I spot each complete successful AI correct now. AI is everywhere, successful everything, and yet immense numbers of group opportunity they dislike it. We spot it, successful nan Decoder inbox, successful nan comments connected The Verge, and connected our videos connected YouTube. Everyone says they dislike it — and yet numbers don’t dishonesty astir really galore millions of group are utilizing it and apparently deriving immoderate benefit. 

It’s a large contradiction and difficult to unpack. But Prashanth was beautiful consenting to get into it pinch me. I deliberation you’ll find his answers and his penetration very interesting.

Okay: Prashanth Chandrasekar, CEO of Stack Overflow. Here we go.

This question and reply has been lightly edited for magnitude and clarity.

Prashanth Chandrasekar, you’re nan CEO of Stack Overflow. Welcome to Decoder.

Wonderful to spot you again. It’s been a basking minute. It has been 3 years since nan past clip we spoke, truthful it’s awesome to spot you again.

I should person said invited backmost to Decoder. You were last connected nan show successful October 2022. One period later, ChatGPT launched. 

[Laughs] That was an interestingly timed interview, correct earlier nan world changed.

Right earlier nan world changed. Software improvement is surely nan point that has changed nan astir since AI models person hit. There are a batch of caller products successful your beingness to talk about, and there’s what Stack Overflow itself is doing successful nan world of AI. So I want to talk astir each of that.

But first, return maine backmost to that moment. We had spent an full speech successful 2022 talking astir organization and moderation, really you were going to build a chimney of group learning to code, and learning to usage Stack Overflow. That was a large portion of our conversation. The technologist pipeline, some them learning to constitute package and being a portion of nan package improvement community, was very overmuch connected your mind. And then, each of package improvement changed because of nan AI tool. Describe that infinitesimal for maine because I deliberation it contextualizes everything that happened afterwards.

It was decidedly a very, very astonishing moment. It wasn’t an unexpected infinitesimal successful galore ways because present came this exertion that evidently immoderate group knew astir but not successful a measurement that captured everybody’s imagination utilizing this beautiful interface. We were successful nan mediate of wrapping up our almanac year, and astatine that point, we were reasoning astir our priorities for nan adjacent year. 

It became very clear what we needed to attraction connected because this was going to beryllium this very, very immense alteration to really group devour technology. Welcome to technology. It’s perpetually changing, and I deliberation this activity particularly is wholly unprecedented. I don’t deliberation location is immoderate benignant of affinity aliases anterior activity that I could look to, including nan unreality and possibly nan internet. I don’t deliberation we’re still afloat consuming what this is astatine nan moment.

So, we went into what is nan balanced of a codification reddish business wrong nan company. It was an existential moment, particularly for our nationalist level because nan superior job, if you will, is each astir making judge group get answers to their questions. Now, you person this really, really slick earthy connection interface that allows you to do that astatine a moment’s notice. We had to benignant of shape our thoughts, and what I ended up doing was carving retired 10 percent of nan company’s resources to specifically attraction connected a consequence to this. 

We group a circumstantial day to respond successful a meaningful fashion, truthful we said nan summertime of 2023. I was going to spell speak astatine nan WeAreDevelopers Conference successful Berlin, and I efficaciously told nan company, “We’ve sewage six months to spell and nutrient our response.” At least, it would beryllium our first consequence because this is going to support iterating.

That’s really we mobilized nan company. We acknowledged it was a codification reddish moment, we carved retired a squad of 10 percent, truthful astir 40 group aliases truthful since we were a somewhat medium-sized company. Then, we sewage to work. That was nan moment.

Take maine wrong that room. Very fewer group ever get to send nan codification reddish memo, right? This is not a point astir group ever get to do. Maybe you deliberation astir doing it, but nary one’s going to publication your memo. Everyone has to publication your memo. You’re nan CEO. 

Take maine wrong that room wherever you said, “Okay, I person identified an existential threat to our company. People person travel to america for answers to package improvement questions.” Again, nan past clip you were connected nan show, you were talking astir nan thought that location were objectively correct answers to package improvement questions and that nan organization could supply them and ballot connected them. Well, now you’ve sewage a robot that tin do it and tin do it arsenic overmuch arsenic you want, arsenic agelong arsenic you want. There are devices for illustration Cursor AI and Cloud Code that tin tally disconnected and do it for you. 

So you’ve sewage each that, and you say, “I request to return 10 percent of nan company.” I’m funny really large nan institution is. I cognize there’s been immoderate changes, but 10 percent of nan institution is 40, 50 people. How did you place and say, “This is nan infinitesimal I request to propulsion these group successful nan room. I’m making this decision, and nan correct reply is 40, 50 group are going to group speech their clip to present maine a scheme by this clip adjacent year?”.

The small heart has travel from a mates different experiences. My acquisition correct earlier this was astatine Rackspace successful nan unreality services space. The business I was moving astatine Rackspace was really astir responding to Amazon Web Services arsenic a unreality exertion threat. I was connected nan squad that yet built that business from nan crushed up, and it was efficaciously 10 percent of Rackspace’s organization that went and created that. So, I had immoderate believe seeing and responding to a disruptive threat that you encounter. It was my move now to put that into mobility astatine Stack by appointing personification for illustration myself erstwhile I was astatine Rackspace to do precisely nan aforesaid thing. 

The different information constituent goes each nan measurement backmost a mates decades aliases much erstwhile I was successful business school. My professor was Clayton Christensen, and he wrote nan book The Innovator’s Dilemma. I’ve ever thought astir that successful nan discourse of technology. In technology, it is simply a very accordant taxable that each truthful often, you will person disruptive threats, and there’s a very circumstantial measurement successful which you request to respond to that. It’s very overmuch astir really history suggests you should carve retired an autonomous squad that has very different incentives and tin prosecute things successful a very different measurement comparative to nan remainder of your business. 

And remember, Stack Overflow is really 2 parts. We person our nationalist platform, which is this large web disruptor and which we should talk astir much broadly successful regards to nan internet). The different broadside is nan endeavor business, wherever we are serving ample companies pinch a backstage type of Stack Overflow. Thankfully, group proceed to spot worth successful having a knowledge guidelines that’s very accurate. Increasingly complete nan past fewer years, it’s really moreover go much valuable because you request really awesome discourse for AI agents and assistants to work. I’ve sewage plentifulness of examples, truthful we tin talk astir that. 

So, that’s wherever that consequence came from. I had been done it successful a mates different dimensions anterior to that. And conscionable successful position of really I communicated to nan team, nan memo was really for illustration a bid of memos. Every Friday, I nonstop a institution email. I conscionable sent 1 correct earlier I sewage connected here. I americium beautiful transparent successful those:  “Here’s what’s connected my mind, here’s what we should beryllium doing, present are immoderate awesome things that happened, present are immoderate group who demonstrated halfway values.”

I’ve done that religiously for… I’ve been astatine nan institution for six years, and I do that each Friday. The squad fundamentally knows what’s connected my mind, and truthful it wasn’t 1 large memo to activate it. It was a bid of emails starring up to this infinitesimal saying, “here’s what we’re going to be, we’ve sewage to respond to this, here’s what we’re reasoning astir now,” and truthful connected and truthful forth. This went connected until I could put nan emblem canvass down and say, “Hey, by nan WeAreDevelopers Conference, we person to nutrient a meaningful consequence connected nan nationalist level arsenic good arsenic connected nan endeavor beforehand because it’s a awesome opportunity to merge AI into our SaaS application, which is simply a different vector.” Hopefully that helps.

Did you really type nan words “code red?”

I deliberation I decidedly utilized “disruptive.” I utilized “existential moment.” I utilized each those things, but I don’t cognize if I utilized nan nonstop words, “code red.”

[Laughs] I conscionable deliberation astir that infinitesimal wherever you’re like, “All right, I’m going to deed nan C and nan O… I’m saying these words, it’s happening.”

It was very clear. We person a very circumstantial connection cadence pinch nan company, for illustration galore others. The reside and seriousness of what we were moving connected was very evident to people, particularly erstwhile you carve retired resources and return group distant from definite teams. People are going to ask, “Wow, what astir my staff?” Here you go, there’s nan reason. So, it becomes very obvious.

How did you make those decisions to propulsion group away? How did you determine which people, really did you determine which teams? Those are each trade-offs, right?

Yeah, nary doubt. I deliberation this is simply a difficult problem to solve. You surely want very talented people, but I deliberation you want nan types of group who are consenting to break solid aliases spell against nan atom and not beryllium encumbered by humanities norms. I very specifically picked a operation of people. The group who are starring it were newer and came from nan extracurricular of nan institution because remember, we were going done a transformation. 

I joined a institution that was engineering-led successful 2019 and each astir this nationalist platform, and we were transforming into this product-led organization. We specifically appointed a newer personification who had travel from nan outside, was willing successful building highly innovative, accelerated iterating products, and had nan DNA and nan thrust to do it.

I besides personally stayed overmuch person to it. I, successful fact, ran merchandise for an interim play of clip pinch that personification reporting straight to me. That was different measurement to enactment very, very adjacent to what was happening connected nan crushed until nan existent launch. The remainder of nan squad was a operation of very talented engineers, designers, and group who had discourse of really nan tract worked successful nan past and who could supply america pinch each nan unlocks that we needed.

I deliberation astir Stack Overflow successful what are astir apt 2 reductive position successful this context. You person inputs, you person outputs. The inputs are users answering questions. The outputs are nan answers to those questions that group travel and hunt for. There’s a full organization that makes that strategy run. The package level manages that organization pinch nan arena moderators, but it’s really inputs and outputs. There are group who are asking questions and group who are answering questions. 

Both sides of that are profoundly affected by AI. I deliberation we person travel to nan unfastened web portion of nan conversation, wherever nan input broadside is being flooded by AI generated slop. In 2022, you had to ban AI-generated answers successful Stack Overflow. Then, connected nan output side, nan expertise for AI devices to proviso nan answers is overwhelming. 

So, let’s conscionable break it into 2 parts. How did you deliberation astir nan input side, wherever there’s a flood of group saying, “Oh, I tin reply these questions faster than ever by conscionable asking ChatGPT and pasting nan reply in. Maybe that’s not bully enough, but I tin conscionable do it.” Then, really did you deliberation astir nan output side?

We noticed 2 things correct retired of nan gate. One was nan number of questions that were being asked and answered connected Stack went done nan tile because group started using, to your point, ChatGPT to reply these questions. That fueled this spike, which is benignant of counterintuitive, but I deliberation group conscionable felt like, “Wow, I tin crippled nan system, truthful fto maine spell do it.” Very quickly, we had to beryllium highly shrewd, and our organization members are astonishing astatine figuring retired what’s existent and what’s not. They were capable to telephone retired very quickly that these posts were really ChatGPT generated. That’s benignant of what initiated nan ban, which we wholly supported and still support, by nan way. You still cannot reply immoderate of nan questions connected Stack Overflow pinch AI-generated content.

The logic for that, Nilay, is because our proposition is to beryllium nan trusted captious root for technologies. That’s our imagination for nan company. So for us, it’s each astir making judge that location are only a fewer places wherever you tin spell and not woody pinch AI slop, wherever a organization of experts person really voted and curated it truthful you tin spot it for various purposes. On nan input side, it made consciousness to do that, and we proceed to do that. 

Fast-forward a small spot to now, and we person created each sorts of caller introduction points onto nan site, moreover though we’ve had precocious standards to inquire a mobility connected Stack Overflow. We just launched our AI Assist feature into wide readiness earlier this week, and it’s been ace breathtaking to watch really users are utilizing that. It is efficaciously an AI conversational interface grounded connected our 90 cardinal questions and answers. 

Then, there’s nan expertise for group to inquire subjective questions, going backmost to our past speech 3 years ago. Now group are capable to inquire open-ended questions because there’s a spot for Q&A, which is nan canonical reply to a question. There’s besides a spot for chat and speech because there’s truthful overmuch changing. It’s not for illustration each nan answers person been figured out, truthful let’s conscionable make judge that group person an expertise to do that. That’s aligned pinch our ngo of “cultivating community,” which is 1 of nan three parts of our mission. The different ones are “power learning” and “unlocking growth.” So, we person done each these things to make judge that we’re not restrictive connected nan introduction constituent and nan question-asking experience.

The different point connected nan reply broadside is that we realized it’s very important to spell wherever nan personification is spending time. Now that nan world has changed and group are successful truth utilizing Cursor AI and GitHub Copilot to constitute their code, our extremity is to beryllium nan captious root for technology. So let’s show up wherever our users are. We’veactually go a batch much headless. 

For example, we precocious launched MCP servers for some our nationalist level and our endeavor product. What group are utilizing our level to do now is to not only invoke those MCP servers — let’s opportunity if they’re penning codification successful Cursor and want to cognize nan quality betwixt type 1 and type 2 — but to besides to beryllium capable to constitute backmost to our level consecutive from Cursor if they want to prosecute and get a deeper answer, which is very unsocial successful nan industry.

So, that’s been our merchandise principle: conscionable spell anyplace nan personification is. But ultimately, we conscionable want to beryllium that trustworthy captious root for technologists, whether it’s wrong companies aliases extracurricular companies.

How do you monetize successful a world wherever you’re headless, wherever you’re conscionable different database that someone’s querying from Cursor? How does that make you money?

We make money chiefly successful 2 ways. We person a 3rd way, thankfully, but nan 3rd measurement is nan smallest part, truthful I’ll commencement pinch nan biggest. Our endeavor business, what we telephone Stack Internal, is now utilized by 25,000 companies astir nan world. Some of nan world’s largest organizations, banks, tech companies, and unit companies usage this merchandise to stock knowledge internally. Increasingly, they’re now capable to usage that trustworthy knowledge to powerfulness their AI assistants and AI agents to spell do various things. 

A bully illustration of this is Uber, which is simply a customer of Stack Overflow Internal and it has Uber Genie. It has thousands of questions and answers connected our platform. Uber Genie plugs into that contented done our APIs, and past it’s capable to spell into things for illustration Slack channels to automatically reply questions and thrust productivity truthful that you’re not bothering people. It’s rooted successful nan organization’s knowledge connected our platform.

So, nan endeavor business is our superior business. The 2nd business is our information licensing business, which we really built only complete nan past mates of years. One of nan things we besides noticed was that a batch of nan AI labs were leveraging our information for LLM pre-training and post-training needs, on pinch retrieval-augmented procreation (RAG) indexing, truthful we put up a full bunch of anti-scrapers. We worked pinch third-party companies, and very quickly we sewage calls from a batch of them saying, “We request entree to your data. Let’s activity together to formally get access.” We had to do that, and now we’ve struck business agreements pinch each azygous AI laboratory that you tin deliberation of, each unreality hyperscaler that you tin deliberation of — companies for illustration Google, OpenAI — and moreover partnerships pinch Databricks and Snowflake, moreover though they’re not doing LLM pre-training. That’s been our 2nd business much recently. 

And nan 3rd one, which is nan smallest portion of our company, is advertising. I deliberation astir group presume that Stack Overflow is supported wholly by advertising, but it’s only astir 20 percent of our institution revenue. We person a very captive, very important assemblage of developers who walk clip connected nan site, truthful we person ample advertisers that want to get their attraction connected various products. In fact, there’s a batch of title now, truthful they progressively want to do that. That’s really we make money. 

So, successful nan discourse of becoming headless, for america it’s astir our endeavor product. It useful connected a subscription and pinch hybrid pricing. That’s really we make money there. The information licensing is akin successful that if group want access, they’ve sewage to salary for that. Then, yes, advertizing is constricted to immoderate of nan largest companies, and they salary america for that. But there’s ever going to be… I would opportunity it’s an “and” versus an “or.” We’re not going to beryllium wholly headless. I deliberation we conscionable want to springiness nan personification nan action to beryllium headless. Plenty of group still travel to nan site, and successful that case, we’re capable to equilibrium that retired pinch these mechanisms.

Do you deliberation caller users are going to travel to Stack Overflow? Stack Overflow is simply a merchandise of nan mobile era. There’s an detonation of package development. There’s an detonation of community. There’s a civilization successful nan worth of building apps and services, and there’s caller tools. Stack Overflow is 1 of nan cardinal gathering points for that organization successful that era. 

New developers coming mightiness conscionable unfastened Cloud Code, Cursor, GitHub, aliases whatever, and conscionable talk to that. They mightiness ne'er really task retired into a organization successful a akin way. Do you deliberation you tin get group to travel to Stack Overflow straight and activity retired answers from different people, aliases are they conscionable going to talk to nan AIs?

I deliberation that for elemental questions… By nan way, erstwhile we saw nan questions diminution successful early 2023, what we realized is that beautiful overmuch each those declines were pinch very elemental questions. The analyzable questions still get asked connected Stack because there’s nary different place. If nan LLMs are only arsenic bully arsenic nan data, which is typically quality curated, we’re 1 of nan champion places for that, if not nan champion for technology. It’s still a very progressive tract pinch a batch of engagement and a batch of monthly progressive usage. 

The questions being asked are rather advanced, I would say. What we’re besides progressively seeing done nan caller mechanisms that we’ve opened up… because to reply your question, we want to springiness group different reasons to travel to nan tract too conscionable getting their answers. So, we person had to broaden our site’s purpose, hence nan ngo of “cultivate community, powerfulness learning, and unlock growth.” 

What we’ve done is unfastened up caller introduction points, caller ways for group to engage. We, for example, unlocked nan expertise for humans to chat pinch each different to get directional guidance. That’s been a very celebrated characteristic connected nan tract wherever group are engaging pinch different experts. For example, we person group asking OpenAI API questions, and they tin spell into nan OpenAI chatroom and prosecute pinch different group who person akin questions, aliases Python experts.

We besides opened up nan expertise for group to show their knowledge pinch challenges, efficaciously for illustration hackathons.  We’ve opened up a full bid of challenges, which are a very celebrated characteristic now. People walk clip to spell and lick these challenges that we post, and that way, they tin showcase their knowing of nan fundamentals, which I deliberation is very important successful position of wherever nan world is going. 

If group are conscionable utilizing vibe coding devices and codification gen tools, companies bringing successful young talent request to cognize that they’re relying connected group who not only took nan shortcut, but besides understand nan fundamentals. We’re 1 of nan fewer places wherever you tin really beryllium that you’ve learned nan fundamentals. So, that’s nan different logic why we’ve opened up these caller mechanisms.

Then, there’s nan 3rd portion of nan mission, which is unlocking growth. There’s going to beryllium a batch of occupation disruption because of each this. If people’s jobs are going to alteration rather dramatically, inferior developers are going to request a home, moreover though I deliberation it’s a shortsighted move by galore companies to extremity hiring them considering you request a pipeline. They’re going to request to link pinch different people, to beryllium capable to progress, learn, and get jobs. Jobs are a very important part. We struck a business pinch Indeed this past twelvemonth to partner connected tech jobs. It’s conscionable to broaden nan scope of our tract truthful that location are galore different reasons different than asking nan questions. They still do, but we besides want to springiness them much reasons to travel to nan site.

This comes to nan large hostility successful each of this. I spot it playing retired successful each kinds of different communities. I spot it playing retired successful our ain comments successful a batch of ways. You want to build a organization of group who are helping different group get better, and that is being disrupted connected each broadside by AI. Communities that are built astir group are beautiful resistant to nan incursion of AI.

This has decidedly happened connected Stack Overflow. Your moderators person fundamentally revolted complete nan expertise to region AI-generated answers arsenic accelerated arsenic they want to. When you partnered pinch OpenAI, a bunch of users started deleting contented truthful it wouldn’t beryllium fed into OpenAI for training, and you had to prohibition a bunch of them. How are you managing that balance? Because if you build communities astir people, I would opportunity nan civilization — correct now anyhow — is that those communities will push backmost against AI very hard.

I would opportunity 1 of nan astir important things that we’ve focused connected (and that I’ve spent clip connected complete nan past fewer years), is this full push and propulsion of really we deliberation astir AI successful nan discourse of our site. Because it’s beautiful clear to america that if we don’t modernize nan tract successful nan discourse of america leveraging AI arsenic an introduction constituent that it’s going to beryllium little applicable complete time. That’s not good. So, we’ve taken a very fierce stance by incorporating AI into nan nationalist level pinch AI Assist, which has been awesome to see. I’ll locomotion you done nan determination connected why we did that. Then, we did nan aforesaid point connected nan endeavor side. 

If I deliberation astir nan personification guidelines astatine Stack Overflow, it’s benignant of for illustration a large nation, right? We’ve sewage 100 cardinal people, and there’s decidedly group connected some sides of nan spectrum. We person thing called nan 1-9-90 rule. One percent are nan hardcore users who person spent a batch of clip pinch their blood, sweat and tears curating knowledge, spending their clip connected nan site, and contributing. Nine percent are doing it successful a mean way, and 90 percent are consuming and mostly lurking. 

We inquire group connected nan tract whether aliases not they’re utilizing AI. Our ain surveys fundamentally say, if you took a look astatine nan Stack Overflow 2025 Developer Survey, complete 80 percent of our organization members are utilizing AI aliases intend to usage AI. Eighty percent. But nan spot level erstwhile they’re utilizing AI is only astir 29 percent. Only 29 percent of our personification guidelines really trusts what’s coming retired of AI, which is really rather due considering wherever we are because location should beryllium skepticism of this caller technology. 

So, there’s enthusiasm to effort it but not to afloat spot it. And pinch this 1-9-90 rule, I deliberation what we person is simply a halfway group of users that are ever going to beryllium nan protectors of nan company’s original mission, which was to create this wholly meticulous knowledge guidelines and do thing more. Then, we person a very ample number of group who are, let’s say, nan adjacent procreation of developers, who are looking to leverage nan latest and top tools. It’s very clear to america based connected surveys and further investigation that they want to usage earthy connection arsenic nan interface to beryllium capable to do this.

It is nan astir meaningful alteration successful position of machine subject development. If you look each nan measurement backmost to object-oriented programming galore decades ago, that wasn’t really specified a immense boom. It didn’t really create this benignant of change. But now, we’re successful this infinitesimal wherever everything’s been unlocked. It’s a immense alteration effort, and we’ve had to determine to respect nan original ngo and support accuracy astatine nan bosom of it. We’re not comfortable utilizing AI for answers, for example, because it will make slop. It hallucinates, hence why nan spot people is low. But why don’t we incorporated earthy connection interfaces truthful that’s nan preferred measurement to engage? So, we ended up doing that, some connected nan nationalist broadside arsenic good arsenic connected nan endeavor side.

That’s been good received by nan immense mostly of users, but location will ever beryllium a vocal number who will push backmost against incorporation. Beyond nan site, there’s conscionable a level of broader interest astir what each this does to jobs, and what’s going to hap if we fto nan feline retired of nan bag. So,there’s that interest also, which is understandable.

Let maine put a good constituent connected that. I deliberation I understand that successful a sharper way. If I americium personification successful your 1 percent who spends a batch of clip connected Stack Overflow helping different people. The logic I reply questions for free connected your platform, which you monetize successful tons of ways, is because I tin straight spot that my effort helps different group turn and that I’m helping different group lick problems. That is 1 very self-contained dynamic. The past clip you were connected nan show, our full speech was astir that move and really you sewage group to participate successful that move and nan worth of it.

Then suddenly, there’s a very clear economical use to nan institution that owns nan database because it’s trading my effort to OpenAI, which is happening crossed nan board. It’s going to do these information licensing deals pinch each these AI providers, they’re going to train connected nan answers that I person painstakingly entered into this database to thief different people, and now nan adjacent procreation of package engineers is going to get auto-complete that’s based connected my activity and I’ve gotten nothing. I’ve heard that from tons and tons of people. I’ve heard that successful our ain community, and I deliberation I person felt that arsenic various media companies person made these deals.

How do you respond to that? Because it feels for illustration you were providing a database that you had to monetize successful immoderate ways, but nan relationship group had was nan value, and now there’s different benignant of economical worth that is possibly overshadowing, recasting, aliases re-characterizing nan relationship that group have.

There are a mates of points there. One is astir this company’s original DNA and why group came together to do this thing. When I joined nan company, I asked a mobility like, “What’s people’s inducement to walk clip doing this?” I asked nan founders, specifically [co-founder] Joel Spolsky, astir this. His constituent was that nan package improvement organization is very altruistic. People conscionable want to thief each different retired because group understand really frustrating… I utilized to constitute codification galore years ago. I precocious picked it backmost up pinch immoderate of nan code-generation tools, which is absorbing to comparison and contrast. I conscionable retrieve really frustrating it was if you sewage stuck connected something. Stack was a immense boon erstwhile it was created to unlock this. It was genuinely retired of that. That was nan reason. 

Even earlier ChatGPT, we besides asked nan question, “Should we incentivize users by paying them? Should we springiness them a monetary benefit?” That wasn’t a precocious inquire by a personification base. We went and researched people. People were not successful for nan money. Plus, it complicates things because really do you judge nan costs for a peculiar JavaScript mobility comparative to a peculiar Python question? It goes down a rabbit hole, which is untenable. So that’s one. What was nan original logic group sewage together; it was astir nan mission.

Secondly, successful position of why we person to do this and if it’s unfair. The superior logic we person to spell down nan licensing way is because nan exemplary of nan net has virtually been turned upside down. I cognize you talk astir this, Nilay, pinch nan “DoorDash problem.” People relied connected nan exemplary of nan net wherever group spell to hunt engines and websites and you monetize disconnected of ads. I really empathize pinch contented sites that are heavy limited connected advertizing because I deliberation astir contented sites’ postulation is down 30 aliases 40 percent, thing for illustration that. There’s this immense seed displacement wherever companies that support these platforms person to… we’re a business ultimately. So, what do we person to do? We person to do what is basal and adopt a caller business exemplary to survive, thrive, and do each nan things.

Thankfully for us, we had an endeavor business, which is independent of each of this. Thankfully for us, we still had nan advertizing business, and ample advertisers still cared astir our community. So, information licensing only felt correct successful position of making judge that we tin efficaciously capitalize connected nan infinitesimal and beryllium capable to put backmost into our organization truthful that group who are location for nan correct reasons saw nan benefits. We’ve invested pinch each these caller features I conscionable mentioned. Whether it’s these caller contented types, challenges, chat, AI Assist, immoderate of these things, they each return resources to spell and build. So, we had to spell and leverage nan costs we received to beryllium capable to spell do that.

Now successful nan future, we whitethorn see different ways. For example, should we salary our users, springiness them a portion of nan information licensing revenue? Perhaps. We ever inquire that question. There are ever ways for america to continue, but this is nan existent setup that we person correct now. It’s balancing a batch of things.

You mentioned that to get to nan information licensing deals, you had to put up a bunch of anti-scraper tools. You had to spell into secondary and tertiary layers of nan stack to get deals from Databricks and different kinds of providers. The AI companies were conscionable scraping your tract before. They astir apt still are. Whether aliases not they’re paying you, they’re astir apt still conscionable going done nan beforehand doorway because each of them look to beryllium doing that. Did you person to say, “We’re stopping you,” and past spell get nan deal? Or did you say, “Hey, we cognize you’re doing this, but you person to salary america aliases we’re going to commencement litigating?”

It’s location successful between. We put up nan anti-scrapers very quickly. We moreover changed nan measurement successful which group received our information dumps. Again, there’s a equilibrium because we ne'er wanted to forestall our organization users from grabbing our information for their morganatic needs, for illustration their schoolhouse projects aliases PhD theses. So, we’ve continued to beryllium unfastened astir our information for our organization members, but they person to beryllium organization members, and location tin beryllium companies looking to commercialize disconnected nan data.

We were very circumstantial astir nan argumentation terms. We put up exertion that prevented group from grabbing it, truthful we knew precisely who is scraping and who’s not scraping. We reached retired to immoderate of those folks and said,, “Look, guidelines down because you’re putting a batch of unit connected our servers by doing what you’re doing, truthful return it easy successful here.”

But I deliberation my characterization of those companies is that they don’t care. Some of them attraction and want to beryllium bully citizens, and immoderate of them perfectly do not attraction and they would for illustration nan smoke. You tin conscionable categorize them. There’s a logic Amazon is suing Perplexity. They told Perplexity to extremity it and Perplexity won’t. The New York Times, arsenic we’re speaking today, is suing Perplexity. Then, location are different players acting successful different ways and striking different kinds of deals. 

Walk maine done 1 of those deals. When you went and struck your woody pinch OpenAI, was it, “We’re going to extremity you, and if you want nan doorway to beryllium unfastened again, you person to salary us?” Or was it, “You cognize this is wrong. We tin return each nan method and ineligible measures, but we should really conscionable get to nan woody correctly?” Walk maine done that conversation.

We were incorporating thing for illustration OpenAI into our product. Remember nan codification reddish business wherever we were astir to denote our AI response. So, we were really utilizing that exertion to do what we had to do to incorporated AI into nan nationalist level and our endeavor product. We had a narration pinch them, and we besides said, “Look, this is not going to work. It’s not tenable, and this is nan caller measurement of working. Maybe we request a caller business statement for you to usage nan data. Let’s really person a conversation.” And in installments to them, they were very partner-centric astir that. I was very impressed by OpenAI and companies for illustration Google that are each very unfastened to engaging connected this taxable and wanted to beryllium responsible AI partners.

They sewage it immediately, moreover earlier we asked them. It wasn’t this big, “Let’s spell person this speech from nan crushed up and warrant why it had to beryllium done.” We conscionable said, “Look, this is what needs to hap because this is simply a caller business model.” We sewage into nan speech beautiful quickly. “What precisely are you looking for? Which format of information do you want to scrape nan content? Do you want bulk uploads? Do you want API calls? What do you want?” 

So, we sewage into that full mix. And punctual you, Nilay, these are recurring revenue-type deals. These are not 1 clip payments. If you want entree and you want continued entree successful nan future, you’ve sewage to support paying, moreover pinch nan humanities data. So that’s really these are group up.

So yes, they were very collaborative partners. But you’re right. There are contradictory players. They opportunity thing and their actions beryllium different things successful position of really they’ve engaged. There are holdouts for judge and group who are not precisely accordant pinch their word, and that’s unfortunate. I deliberation each institution for illustration america has to determine what to do astir that. We’re successful various stages of these conversations pinch group connected really to make sure  we sensibly get them to do nan correct thing.

Now you person to sanction 1 of those companies. Who do you deliberation is holding retired otherwise than their nationalist posture?

I’d alternatively not beryllium specific, but each nan accustomed suspects that you’re covering are nan accustomed suspects that we are encountering. That’s really I would put that. 

Let maine inquire you astir nan recurring gross portion and past I want to get into the Decoder questions because I deliberation they’ll beryllium illuminating aft this conversation. There’s a consciousness that we’ve done each nan pre-training that we’re going to do, right? Scraping nan net is not nan early of these models, and location needs to beryllium immoderate different leap. 

Stack Overflow’s existing corpus of accusation is nan valuable thing. There’s a batch of information. There’s 20 years of worldly successful that database. What’s nan worth of “you person to salary america again to train nan adjacent type of Gemini aliases GPT” and nan worth of “there’s incremental accusation being added to nan existing database?” Because that seems for illustration a clear divided to me.

The measurement we’ve thought astir this is that each exemplary that’s being trained is trained connected immoderate corpus of information. You’re going from GPT X to Y. If you’re leveraging our original information aliases immoderate derivative of that from a anterior exemplary successful nan caller exemplary that you’re training, past you person to salary america for it. That’s efficaciously nan ineligible request for doing that. So, it’s a cumulative aspect. Let’s not hide that. People person to salary for nan cumulative data. It’s not conscionable that it was utilized backmost successful nan day. And yes, comparative to 20 years, 1 year’s worthy of accusation is going to beryllium less, but that’s why you’re getting 20 positive one. That’s nan idea. So, that’s nan measurement nan ineligible statement has been group up.

Is it per year? Is it that each year’s worthy of information is simply a chunk of money? How does that work?

No, it’s cumulative, for illustration nan full corpus: humanities information arsenic good arsenic thing going guardant for nan pursuing year. All that is 1 accumulated information set, and that’s efficaciously charged arsenic one.

So, this year’s information doesn’t get pulled into Gemini 3’s information set, which conscionable came out, right? Every caller mobility and reply successful Stack Overflow since Gemini 3 came retired is not incorporated successful Gemini 3’s training.

Correct.

So you’re benignant of betting that they’re conscionable going to train ever-bigger models. Is that really it’s system successful your mind?

Yeah. And immoderate companies person asked for wide usage cases. There are pre-training usage cases. Even beyond that, you tin leverage nan information successful galore different ways for AI and non-AI usage cases, for illustration hunt usage cases. But correct. There whitethorn beryllium scenarios wherever much larger models are built, and our information is going to beryllium useful for those scenarios. But there’s going to beryllium RAG indexing, post-training needs, each sorts of scenarios. It’s rather absorbing to spot immoderate of nan frontier labs inquire for very circumstantial slices of information that they find useful. 

Remember, there’s not only questions and answers. We’ve sewage nan remark history, nan metadata history, nan voting history, nan history of User A going down this path. So, it’s a batch of fantabulous discourse for things for illustration reasoning and being capable to mimic nan quality brain. It’s almost for illustration 1 quality encephalon that’s been documented. 

This, I think, brings maine to nan Decoder questions. You’ve restructured nan company. There’s been immoderate rounds of layoffs. You’ve refocused connected nan SaaS business successful a existent way. I deliberation we should talk astir that. But there’s nan thought that we’re going to train ever-bigger models and that will beryllium nan increasing portion of nan business, versus wanting immoderate slices, versus RAG really being nan early for a batch of these different businesses. You would make different decisions based connected which 1 of those is going to turn faster, and I don’t deliberation anybody knows. Maybe you know. You tin show maine if you cognize aliases you cognize personification who knows, 

But we’re successful a very nascent play for each of this development. How person you system nan institution to moreover retired each of that consequence and beryllium prepared for really group will really request nan information successful nan future? Maybe you know, I don’t deliberation I know.

It’s difficult to foretell clearly. We’ve sewage immoderate superb minds, for illustration Demis Hassabis astatine Google and others, who are coming up pinch nan adjacent generational leap of immoderate nan balanced of transformer exertion is to support going towards this eventual goal, which is AGI.

So yes, you’re right. It’s difficult to cognize precisely what shows up and when. However, nan measurement we are system arsenic a institution is successful efficaciously 2 parts. One portion is nan endeavor business. We person a merchandise team, engineering team, and a go-to marketplace squad focused connected that. The endeavor products business is very clear. Then, nan different broadside of nan location is what we telephone “community products”. That squad focuses connected nan nationalist platform, each nan features that we’ve talked astir truthful far, AI Assist and each nan subjective caller questions, and chat. This is nan organization side. The information licensing business sits successful that group, and truthful they are tied to nan engagement of nan site. So, there’s a virtuous rhythm there.

So, that’s really we’re split. Again, that includes merchandise resources, a mini go-to marketplace team, engineering folks, etc. Also, there’s our organization guidance team, which spends a batch of clip pinch nan moderators to prosecute there. So, it’s divided down nan middle, and our different functions support both.

How large is Stack Overflow today? I cognize you laid disconnected almost a 4th of nan institution successful 2023 because of postulation declines. You’ve built different businesses. How large are you today?

We are astir 300 group aliases so. Yeah, that’s our size.

Do you deliberation nan gross you’re going to spot from information licensing aliases your SaaS business is going to let you to turn again?

We judge so. We’re a increasing company. We’re profitable. So financially, we’re thankfully successful a very bully spot. Now, it’s each astir placing bets connected nan highest maturation opportunities. We judge creating this knowledge intelligence furniture wrong endeavor done our Stack Internal merchandise is simply a phenomenal maturation opportunity because customers are pulling america successful that direction, which is awesome to see.

That’s wherever you’re headed. Those are immoderate of your announcements. I want to talk astir that successful a second, but I’m conscionable focused connected nan early is nan SaaS business for enterprise. The early is information licensing. Are you still seeing declines connected nan nationalist website?

I would opportunity it’s stabilized for a fewer months. I deliberation nan engagement and nan activity connected nan tract are really beautiful stable. The driblet successful questions that I mentioned antecedently were each nan elemental questions, and it seems to person gotten to a spot wherever analyzable questions are being asked. We person a accordant number of group connected nan tract each day. We person thing called a heartbeat. In fact, anybody tin spell cheque it out. If you spell to StackOverflow.com and scroll to nan bottom, you’ll spot really galore users are online astatine nan moment. And truthful you’ll ever spot a very accordant number there. I deliberation it’s difficult to foretell nan future, but surely I deliberation nan worst of it was backmost successful 2023, 2024, for sure..

The mobility I inquire everybody connected Decoder, arsenic you good know, is really do you make decisions. The past clip you were connected nan show, you said that you wanted to beryllium connected nan beforehand lines arsenic overmuch arsenic possible, and you wanted to beryllium informed by group who are connected nan ground. Has that changed successful nan past 3 years? What’s your decision-making process?

Not really. I deliberation it’s very important for leaders and group for illustration CEOs to person nan afloat discourse because you can’t person filtered information. I spent a batch of clip pinch users and customers really knowing what they attraction about. That’s really we moreover decided connected thing arsenic arguable arsenic nan AI Assist feature. It wasn’t evident to say, “let’s spell and build that” if you were not listening connected nan ground. Because if you conscionable perceive to nan header statements, it seemed for illustration group didn’t want to merge AI into Stack Overflow. But nan reality is that many, galore users —  nan 90 percent I mentioned —  wanted a earthy connection interface. That’s what they are comfortable utilizing these days, and that’s what they wanted. So that’s why we decided to do that.

One of nan things that I spot everyplace is that split. You mentioned 1-9-90 before. There’s a very vocal minority. We spot it successful our ain postulation connected TheVerge. We screen AI deeply, and we are told that everyone hates it. I understand why. I understand nan comments. That’s what I’ll say. I get it. 

Then, I spot nan usage numbers. I spot nan postulation connected our sum of AI tools. I spot companies for illustration yours saying that everyone’s utilizing it. There’s immoderate gigantic divided location that is dissimilar immoderate different I deliberation I’ve ever encountered covering exertion complete nan past 15 years. Everyone says they don’t for illustration it, and past they’re utilizing nan hellhole retired of it. 

The only different 1 I tin deliberation of that is somewhat comparable is really group consciousness astir Adobe. Everyone uses nan devices and everyone’s huffy astatine nan Creative Cloud subscription fee. It’s fundamentally nan only comparison I have. It’s not a bully one, it doesn’t representation one-to-one, but it’s arsenic adjacent arsenic I’ve travel to that split. 

What successful your mind accounts for that divided pinch AI wherever group don’t for illustration it, are very vocal astir not liking it, and past we spot nan numbers and everyone’s benignant of utilizing it anyway?

I deliberation it comes down to that information constituent I shared earlier: that 80-plus percent of our personification guidelines wants to usage AI aliases is already utilizing AI for code-related topics, but only 29 percent of that organization trusts AI. Trust is simply a very heavy word. Why don’t you spot something? You don’t spot thing because you don’t deliberation it’s producing precocious integrity, meticulous answers. You whitethorn not spot it because it whitethorn switch you 1 day, and you don’t for illustration that either. 

But astatine nan aforesaid time, you’re going to beryllium funny astir what’s going to beryllium specified an iconic force. So, you want to support trying and utilizing it and possibly getting amended to ultimately, hopefully, leverage it to your use truthful that you tin beryllium applicable arsenic an individual developer aliases to spell a batch faster successful nan future.

I deliberation that’s astir apt nan reason, particularly pinch nan developer audience. I deliberation they’re a very discerning, let’s opportunity analytical audience, and they tin beryllium prickly if things are not deterministic pinch nan measurement it has been for a very agelong time. This is simply a very probabilistic benignant of technology. It’s almost for illustration going to a casino and utilizing a roulette wheel. You’re going to get a different reply each time. It’s not needfully comforting for personification who’s penning very circumstantial codification and looking for very circumstantial outcomes. 

I deliberation group will get utilized to that complete time. It is simply a mind displacement alteration for group penning software. That whitethorn beryllium nan logic why group are intrigued, because it is truthful powerful arsenic a technology. Don’t get maine wrong. We usage vibe coding each complete nan spot astatine Stack. All nan features I mentioned to you, our designers and merchandise managers vibe coded it first to show it and get personification feedback earlier we went and built it. We’ve embraced these devices internally for that benefit. So, location will beryllium ways successful which you consciousness comfortable utilizing it, but I deliberation that’s nan halfway reason.

I really want to talk astir that dependency. You cognize that it’s not trustworthy, but you are building products pinch it. You are building products to alteration it. The large announcement this week is Stack Overflow AI Assist. You’ve talked astir it respective times passim this conversation. You’re betting that this is what group want, right? 

You’re betting that an AI-powered instrumentality connected Stack Overflow will thief much people. Maybe that point is going to hallucinate for illustration crazy and springiness group nan incorrect answer. Maybe it’s going to, I don’t know, tell Kevin Roose that it loves him connected nan beforehand page of The New York Times. I conscionable for illustration teasing Kevin. Hi, Kevin. 

How do you make that stake erstwhile you cognize that nan users don’t spot it, but you still person to rotation retired nan devices because that’s wherever nan manufacture is going?

We judge we’ve really unlocked a very important facet of that spot rumor and responded to it. Our AI Assist characteristic is simply a RAG positive LLM solution. Effectively, it provides an reply that first goes retired into our corpus of tens of millions of questions and answers. We person 80 to 90 million. Those are first utilized to nutrient a response. If they don’t, there’s a fallback action wherever it goes and leverages our partner OpenAI, for example, to beryllium capable to spell and nutrient trustworthy knowledge from different parts of nan web. So, it first searches done our trusted, attributed knowledge base. It produces nan links truthful group tin spell down that way and study much astir it. Attribution is very important to us, right? 

That’s really we are navigating nan mirage element., We’re perpetually testing it, and it’s not perfect. There will ever beryllium improvements. But we’re besides looking astatine wherever nan world’s headed, and if these models proceed to get better, past we should use from those improvements. Ultimately, we should person nan champion solution because you’re getting grounded quality discourse positive nan LLM strengths.

I deliberation nan point that I’m astir willing successful is nan religion that nan models will proceed to get better. I’m not 100 percent judge that’s true. I’m not judge that LLM exertion arsenic a halfway exertion tin really beryllium intelligent. As you’re saying, group are very attracted to nan earthy connection constituent of these models and nan interface displacement that’s happening. There’s nan level displacement that we’re each going to create package pinch earthy connection aliases fto nan LLMs logic and fundamentally self-prompt themselves into an answer.

There’s thing location that seems risky. Are you perceiving that consequence today? Are you factoring that successful aliases are you saying, “This is wherever we’re astatine now and we person to proceed until thing changes?”

I deliberation nan first point astir nan LLM betterment level is that I’m pinch you. It’s difficult to cognize really things are going to improve. When you conscionable deliberation astir nan past six months and that it’s plateaued, boom, present comes Gemini 3. Again, we’re proud partners of Google. It is simply a seed shift. It blows each different exemplary retired of nan water. Now, we’ve sewage a codification reddish business successful different companies, different LLM competitors. So, it’s —

Sam Altman did type nan words “code red,” by nan way. I want to opportunity that. So that’s very good.

Perhaps that was nan measurement for maine to spell backmost successful nan day. The constituent is that it is astonishing that you’re capable to nutrient that benignant of a leap erstwhile things person seemingly plateaued. I don’t know. I can’t foretell that because these folks for illustration Demis are deep, heavy successful nan subject. So, that’s true, but there’s besides going to beryllium different innovations that we are not privy to. Like I was explaining previously, transformers were evidently a immense improvement successful this space. There whitethorn beryllium thing that these AI investigation labs travel up pinch that we’re not moreover alert of that’ll push things. Ultimately, we cognize that nan compounding effects are very real. We’ve sewage unlimited compute, highly powerful chips, and GPUs that are now lowering their costs.

I was astatine AWS re:Invent this week wherever they talked astir nan Trainium chips, withTrainium3 and Trainium4 being built out. There’s going to beryllium conscionable nan proliferation, and past you’ve sewage entree to data, which we’ve already talked about. So, erstwhile these things harvester and compound together, it’s going to nutrient very magical outcomes. I deliberation that’s nan belief and why my ain assumptions are rooted successful nan truth that it’s going to amended overall.

The logic I’m asking astir this successful nan discourse of nan devices you’re building and everyone utilizing it but only 29 percent of group trusting it is because you’ve sewage to bring that number measurement up to scope nan returns that each institution investing successful AI is trying to reach, including yours. I don’t cognize if nan halfway exertion tin do that. I don’t cognize if you tin stack a bunch of technologies to do that. 

But I do cognize that 1 type of nan early of package improvement is everything vibe coding each nan time. Another type of nan early of package improvement looks for illustration penning intensely agelong prompts for models that are pages and pages themselves, which seems ridiculous to maine but possibly that is nan future. 

Another type looks like, “We return to humans astatine nan cutting separator of package development, and they are co-developing pinch an AI exemplary and past possibly asking Stack.” That feels for illustration a richer, much absorbing future, but it’s unclear to maine wherever we are connected that spectrum aliases really that moreover plays out.

I deliberation that we person not only a bird’s oculus position successful nan discourse of our nationalist organization pinch this 29 percent information point, but our endeavor customers springiness america a clear position connected wherever they are. The ROI mobility is being asked very heavy wrong companies, and I deliberation 2026 is going to beryllium nan twelvemonth of rationalization. If 2025 was nan twelvemonth of agents — wherever each tool’s being tried retired wrong these companies — there’s a very unfastened scenery for CTOs to spell and bargain and trial various tools. So, I deliberation it’s been a tremendous clip for immoderate of nan companies building these tools. 

But 2026, nan CFOs will unit with, “Okay, productivity improvements person to travel from these. We’re going to prosecute little group arsenic yearly readying happens.” There’s going to beryllium tremendous unit successful nan strategy to beryllium what nan existent worth is. Everybody astatine nan elder level that I’ve talked to acknowledges that this is simply a large shift, and they each are leaning into it beautiful hard. They’re each waiting for nan improvements. I deliberation astir companies will opportunity that they person seen improvements successful nan mini groups wherever they’ve tested these tools, but that’s a self-selecting group because they’re enthusiasts. They will spot awesome productivity gains, which is astir apt true, but there’s an absolute drop-off successful productivity arsenic you deliberation astir nan take crossed nan enterprise. Probably for nan aforesaid reasons, by nan way. You’re telling labor to usage devices which whitethorn put them retired of a job, truthful why would they want to do it? Or much fundamentally, if these devices are not cleanable and they’re hallucinating and will beryllium held accountable, that’s not bully either.

Then of course, there’s nan process change, nan mindset change, that you person to wholly alteration your workflows of really you work, each nan endeavor alteration guidance work. That is why our solution, Stack Internal, is building this quality curation layer, aliases knowledge intelligence layer. With nan MCP server connected top, nan knowledge guidelines successful nan middle, and past our expertise to ingest knowledge from different parts of nan company, we tin create these atomic Q&A that are highly adjuvant to guidelines your endeavor knowledge successful these AI agents. That is that solution. That’s why we’ve really gone difficult astatine producing that.

We’ve seen a really beardown consequence from our customers. We person immoderate of nan world’s largest companies leveraging, testing, and building this pinch us, for illustration HP, Eli Lilly, Xerox, each these companies. It’s been astonishing to spot them gravitate to america because they want to fulfill that ROI constituent that you’re making. What are nan gaps? It’s spot again. So, they want this spot furniture done a institution for illustration Stack that they tin really insert successful betwixt their information and their AI tools.

When you opportunity nan property of rationalization, what I perceive is that you deliberation nan bubble’s going to popular successful 2026.

I deliberation nan exuberance successful conscionable trying retired various devices and these unlimited AI budgets will yet travel to roost. I’m not judge astir nan bubble bursting. There’s decidedly going to beryllium corrections on nan way. There’s nary mobility astir it if you look astatine history. The number of vendors that are trading into these companies. What I’m very amazed by is that there’s akin functionality. There’s 4 of them being tested wrong these companies. 

All these companies whitethorn person gotten to $100 cardinal successful yearly recurring gross (ARR), but astatine immoderate constituent there’s going to beryllium churn erstwhile nan CTO decides, “You cognize what, I’m only going to possibly usage 1 and possibly a 2nd 1 arsenic a backup.” No different from nan cloud. If you deliberation astir nan multi-cloud world backmost successful nan day, group didn’t person 3 clouds retired of nan gate. Now you person possibly 1 superior unreality and 1 secondary cloud. This is different, but astatine nan aforesaid time, you’re going to person for illustration 4 different vibe coding devices successful my opinion.

When you deliberation astir Stack Overflow arsenic being that spot layer, that’s nan worth add, that’s what you tin possibly complaint a premium for complete time. You’re still limited connected a instrumentality that only 29 percent of your users trust. And I cognize you’re talking astir RAG and that your different systems are doing that. How do you deliberation you bring that number up pinch Stack Overflow? Is that imaginable for you to do aliases does nan ecosystem person to do it for you?

I deliberation it’s an ecosystem constituent mostly speaking because that is much a reflection of what group person entree to beyond Stack, right? They’ve sewage entree to each these different options. What we tin attraction connected is being nan astir captious root for technologists. So for us, it’s astir making judge this contented is fantabulous and precocious quality, but that it’s besides a awesome spot for group to cultivate community, link pinch each other, study and turn successful their careers.

I deliberation nan measurement we tin do it is done nan truth that we are moving pinch each these large AI labs and nan truth that our trustworthy knowledge that’s been painstakingly quality curated is going to travel into these LLMs, which yet nutrient trustworthy answers. We’re benignant of 1 furniture behind, but that’s wherever we operate. We run successful that spot furniture aliases nan information layer, if you will, successful nan discourse of LLMs. So, that’s our indirect publication to that 29 percent.

Prashanth, this has been a awesome conversation. You’re going to person to travel backmost sooner than 3 years adjacent time. What’s adjacent for Stack Overflow? What should group beryllium looking for?

Our biggest attraction will beryllium making judge that we build this endeavor knowledge intelligence furniture for companies to genuinely usage AI agents successful a trustworthy way. We are very, very excited astir our Stack Internal merchandise connected nan endeavor side, which we launched a mates of weeks agone astatine Microsoft Ignite, arsenic good arsenic helping our organization users link pinch each different to really study arsenic good arsenic turn their careers connected nan nationalist platform, arsenic I’ve mentioned throughout. 

There are going to beryllium truthful galore avenues and caller introduction points, for illustration our AI Assist, our subjective content, chat, and different things that group hopefully find useful arsenic things alteration astir them very rapidly. They tin beryllium portion of this astonishing organization and thief each different out. So those are nan 2 focuses, endeavor arsenic good arsenic our nationalist community.

All right. Well, erstwhile nan bubble pops adjacent year, we’re going to person you travel backmost and we’re going to opportunity you predicted it.

Thank you, Nilay. I admit it. 

Questions aliases comments astir this episode? Hit america up astatine decoder@theverge.com. We really do publication each email!