1. Podcasts
  2. The Changelog: Software Development, Open Source
  3. A seminal moment in computing (Interview)
The Changelog: Software Development, Open Source

20 Nov 2026 · 1h 13m · 19 chapters

Episode artwork

A seminal moment in computing (Interview)

In short

The “seminal moment” of agentic AI reshaping software engineering, plus WorkOS’s internal AI coworker Atlas and the broader shift toward identity, permissions, and governance for agents. The episode also includes sponsor segments on observability for ClickHouse, speech-to-text from Deepgram, and trust/fraud signals plus agent sandbox governance.

Guests

MG (Michael Greenwich), founder and CEO of WorkOS. Background: engineering-led company building enterprise infrastructure around identity/authentication and integrations; WorkOS Labs builds frontier products. He describes WorkOS’s products including Radar (fraud/abuse detection), Vault (encryption), Pipes (integrations/token management), and WorkOS Studio (internal analytics/chat tool that became a Slack-focused agent concept). Also featured: Mike Shi, head of product for observability at ClickHouse; co-founded HyperDX (acquired by ClickHouse).

Key claims

Agents will increasingly “choose” tech stacks; companies are reinventing their “OS” for software engineering; the Bay Area ecosystem rapidly diffuses agent-building techniques; agent adoption requires identity/authorization and bounded execution.

Notable examples

Atlas (AI teammate learning/automation system; Slack-thread coworker; sub-agents for events and analytics); WorkOS “off.md” enabling agents to sign up without impersonating humans; ClickHouse Observability ingesting OTLP and managing scaling/schema/HA; Coder governance for isolated, auditable compute environments.

Written by AI. May contain mistakes.

Chapters

4:48

The Impact of AI on Software Engineering

Exploration of the evolving role of AI agents in software development.

“Well, friends, I'm back with one of my favorite people in the world.”
14:01

Engineering-Led Culture in Tech

Explore how engineers shape products and manage complexity in tech companies.

“are the ones shaping things, talking to users, driving them forward end to end.”
22:00

The Seminal Moment in Computing

Discuss the rapid changes in tech and personal experiences in this fast-paced era.

“I'm trying to find the tweet because I thought it was pretty interesting that DHH of all people, as you may know, had said.”
28:00

Exploring the Frontier of Development

The speaker describes their experience with coding and the balance between work and leisure.

“um you know I I end up uh for my agents running them typically in fast mode because I don't have a big time interval where I'm like totally focused on it and I'll parallelize stuff and build stuff.”
29:00

The Concept of the Frontier in Tech

Discussion on the evolving nature of technology and the concept of 'frontier' in development.

“That is akin to this being a frontier moment we're all in.”
35:48

Identity and Agents in Technology

The conversation shifts towards the acceptance of AI agents and their identities in the tech landscape.

“Let's just say hours, maybe days, potentially.”
38:20

Purpose of Innovation in Technology

Exploration of the reasons behind technological innovations and their societal impacts.

“The reason to embody it in something like Atlas is not because it makes the AI better.”
42:00

The Impact of AI on Society

Explore the transformative effects of AI and its potential future impacts on humanity.

“I mean, technology has always been the thing that's let us live longer, be healthier, connect better with people.”
44:40

Learning with AI Tools

Discuss the benefits of AI tools in learning and personal development.

“Like, you know, we can cook food because of fire.”
47:00

Optimism vs. Risks in AI Development

Examine the balance between the optimistic potential of AI and the risks associated with its development.

“But every day, every month, to me it feels like the optimistic case in terms of what this will enable is more likely.”
Show all 19 chapters
48:23

Optimism vs. Risks in AI Development

Examine the balance between the optimistic potential of AI and the risks associated with its development.

“It's like, it's just kind of paralyzing waking up and just being like, oh man, this is, you know, it's like being on a rocket ship that's going to go light speed.”
51:15

AI as a Catalyst for Change

Discuss how AI serves as a catalyst for industry transformation and personal creativity.

“And I love the dovetail you just took us on because I totally agree with that.”
56:04

The Evolution of Accessibility in Computing

Discussion on how computers are becoming more accessible to people.

“Let's just start, start the doing, you know?”
57:24

The Challenge of Product Market Fit

Exploration of the complexities in achieving product market fit in today's landscape.

“And is what I do actually valuable to the world?”
59:30

Finding and Addressing Pain Points

Insights on identifying pain points as a strategy for entrepreneurial success.

“one way to do it is to go after something you've had a personal experience in, something that you had like an insight into something in the market, some pain.”
1:02:43

The New Age of Non-Technical Founders

Discussion on the shift in entrepreneurship, emphasizing the importance of agency over technical skills.

“There'll be some thing that that person wants to be done better.”
1:05:38

The Future of Software Development

Predictions on the future of software development and its accessibility to the masses.

“That's one of the things you have to learn.”
1:10:02

Closing Thoughts and Promotion

The guests share final thoughts and discuss the Atlas project.

“Thank you for your wisdom you're sharing here on the, on the show, your anecdotes and hypotheses and all these things.”
1:11:01

Key Insights from the Episode

Recap of significant points discussed in the episode regarding software development.

“Well, friends, what stays with me from this episode is MG's point that for the first time, we can bring computers closer to people instead of dragging people closer to computers.”
A man walking through an airport, asking his phone a question

Ask about this episode

Get the gist, explore an idea or ask about something said.

Claude

On The Changelog: Software Development, Open Source, what are the key points of this episode?

Searching your listening

Transcript

Automatic transcript. May contain errors.

0:00What's up? Welcome back. This is the changelog. My name is Adam Stukowiak. I'm the editor in chief here at ChangeLog and obviously the host of this podcast. You know what? Starting a company has never been easier, but finding product market fit has never been harder. My friend, Michael Greenwich, MG, founder and CEO of WorkOS is back on the pod. We get into Atlas, the AI coworker WorkOS built for itself and then launched. Why agents are becoming the ones who choose our tech stacks for us. and why we all secretly work for the same company. MG and I get into the details of this moment feeling closer to the invention of fire than to the invention of the iPhone, why great founders are heat-seeking missiles for pain, and why calling yourself a non-technical co-founder is cope.

0:47And of course, a massive thank you to my friends and my partners at fly.io. Computers for agents, from sandboxes to production, this is what I do. I give my agents their own hardware-isolated Linux computer, and I deploy what builds without changing platforms. It's magic. Learn more at fly.io. All right, let's do this.

1:16What's up, friends? This episode is brought to you by ClickHouse and their launch of ClickHouse Observability, a fully managed observability service for SREs and their agents. You send your open telemetry data, they handle the rest. I'm here with Mike Shi, head of product for observability at ClickHouse. Mike co-founded HyperDX, which ClickHouse acquired. So he's been building this stack for years. So Mike, is this launch ClickHouse for a whole new kind of team? Yeah, I think ClickHouse has appealed a ton to the builders. And I think that's still going to be that core persona for us. those are building applications and ClickHouse is a great tool for that.

1:56However, I think one of the things that we're trying to allow for are for teams that maybe don't want to build every single bit of their stack, including their observability stack. How do we make sure that the teams are super focused on building their next application? Don't have to worry about, hey, how exactly am I going to be scaling and tuning our ClickHouse cluster for our observability workload? And And that's where ClickHouse Observability really comes in, in terms of taking that burden off these teams, getting started becomes, you're in the OTLP, fantastic, plug into our endpoint, we'll take care of the schema, the scaling, the experience on top, the eval MCP tooling, so that agents can work really, really well with ClickHouse.

2:37We'll take care of all these things. We still get all the benefits and scalability of ClickHouse as well. And for the other things where you do have to worry about all those fine details, you're building something else on top of ClickHouse as well. But when it comes to observability, you don't have to sort of build on top of ClickHouse anymore. You can get something that works really great out of the box. Yeah, out of the box is the part I like. So say I'm that team you're talking about. Open telemetry, it's already flowing and I do not want to babysit yet another database. What actually comes off my plate?

3:06Yeah, so ClickHouse observability is going to basically remediate a lot of the sort of operational concerns when it comes to wanting ClickHouse for observability. So things like, for example, hey, like, how do I make sure my OTEL collectors scale with our telemetry as we send more or less data? That's something that, you know, no longer needs to be a concern. How do we make sure that we have a great high availability pipeline between our ingestion endpoint and clickouts so that when we do have a massive burst in telemetry data, we don't overwhelm our clusters at any given point in time and we don't lose any of those messages as well.

3:41So things like you might have a Kafka or something like that for historically, that's something that becomes, again, managed away from your end. Things like, hey, do we have the most updated schema when it comes to best practice for our workload? Is it tuned to the query shapes that my users are running? Am I optimizing my compression codecs and my sorting keys? All those sorts of questions no longer need to be a concern for the team as well, because, again, those are kind of managed by the platform. So it really alleviates some of the biggest problems we've seen when it comes to users getting started with ClickHouse for observability.

4:15It kind of provides that in a sort of out of the box solution for them. So friends, if you're sizing clusters, tuning schemas and babysitting a Kafka pipeline just to see what production is doing, stop the pain. Stop the pain. Use ClickHouse observability, send your open telemetry data to their managed endpoint and start investigating in minutes. It's in public beta right now at half the standard price. Learn more at clickhouse.com slash observability. Again, clickhouse.com slash observability.

5:07Well, friends, I'm back with one of my favorite people in the world. I say that often, but I have a lot of favorite people, Mike. When you're on that list of favorite people, most people who know you call you MG. And you're the founder and CEO of WorkOS, a company that I've just begun to use the CLI off of. And I'm so stoked about that. Very easy to integrate. But I think we're in some crazy times. And I think you know that, too. So let's just begin with the crazy time, the seminal moment that we're in. What do you think about that? I mean, if you go back a year ago, last summer, I would have probably said the same thing.

5:41We were in crazy times. But back then, you know, we were writing code by typing.

5:51AI was more like kind of like super autocomplete building things. Maybe it was building snippets of things still or small features, but not certainly not like what's working on today. and I think agents were something that people didn't really know what it meant you know like we had uh we talked we talked about them as like ai inside of an ie but people weren't really talking about coding agents in the same way like clog code is like a bit a bit like that but um certainly not the agentic software wave that's happening now yeah so that's totally different I mean, that's, I feel like every six to 12 months, we sort of learn something different about the capabilities of what the AI is going to do and what people can build with it and how to use it.

6:40And it's like this instrument. We keep discovering more things around it. And the possibility for impacts in the world gets bigger and bigger every single time. So I suspect if we're sitting here a year from now. Will I say the same thing? You think? We'll probably be saying the same thing. We'll probably be saying it's crazy times. I can see. And we'll look back on it and be like, oh, 2026, wasn't that cute? You know, how early it was. Yeah, I mean, but like the crazy times that I would describe, let me just maybe paint some version of a picture. And maybe you can justify this because I'm sure you can, you should, but there's Atlas, right?

7:21You're an off company. At least from the outside, I would say you're an off company. But the crazy times isn't just simply the autocomplete to full agentic. Now we've changed the operating system of software engineering as we know it. It's that I think that the crazy time is that it seems like almost every company is some version of reinventing the OS, the work OS that they're on to include literally your company, WorkOS. You've got Atlas, your AI coworker. I mean, that's not exactly a shiny object. Maybe some might say it is, but you had some reason to be bold enough to build this thing. And I think every company is building something for themselves.

8:03And you happen to be in a position where you have great distribution, you have great enterprise companies you work with. And so it's probably easy for you to dovetail versions of that into your product line. That's the crazy times. I didn't think that we would sit here and say, let's talk about this agent harness thing, whatever this is, this coworker that you've built for your internal, which I have no idea how you're using it. But to sort of maybe not so much pause product, but to zoom out further and look at the macro, not the micro of the world we're in and leading companies. And you know this well because you're CEO of a very successful company.

8:41Yeah, Atlas is amazing. it's uh set the stage what is atlas i didn't well so yeah so we so we've been building new things i mean always since the beginning of the company we build a lot of internal software we're exploring new things and in the last two or three years we have a bunch of new products we've built like we have um one called radar that's about detecting fraud and abuse kind of blocking fraudulent actors we have another product called vault that's for encryption we have a product called pipes which is for integrations it's pretty sweet like it lets you connect to a lot of different sass services and manage tokens and integrations especially relevant today so we're always building things i think as a as a company work os is very engineering led we're making things for other engineers and you know maybe about a year ago we started building a lot more things using generative code like code generation systems we had an internal product we built called WorkOS Studio, which we actually showcased at our user conference, Enterprise Ready Conference last year.

9:45We decided not to ship it. It was sort of an internal replacement for Glean, sort of for data analysis, could visualize stuff. We could use it to build internal applications, kind of like Retool or something like that. It was a very different beast. It's almost like a variant of ChatGPT that you could talk to and debug stuff internally and ask questions. and our team really loved it um and so we used it more and more and more it became this thing where a bunch of the company was data active users but what we found was that the collaborative surface of it was pretty minimal like you could create these chats and share it with other people share links um the artifact was kind of like somewhere in between document and an application that you could share but our team lives inside of slack like we're a remote company today we have about 130 people or so um where most of those people like the vast majority are remote and so the question was okay how do we take all of that stuff we learned and make it even more valuable inside of slack as a as an environment and we had seen some demos other people had built or things people had built internally we had a couple slack bots that we had already made like we had one that was for doing go-to-market stuff so our sales team could talk to it and do analysis and research We have another one we actually built for customer support for our end customers with a custom rag pipeline and all this stuff.

11:03So we've been building these Slack bot experiences, but said, like, what if we just made this more of a general purpose internal agent that could do anything? You could hook into it. It uses all the stuff we've built, like the identity stuff, uses pipes for the integrations. So we built it. And Atlas is what we call it. Atlas is actually an acronym. it stands for um ai teammate learning and automation system i love it l a s and the idea is you interact with it it's it's like a co-worker so you can tag it in threads you can tag it to do anything with you um and it learns as you use it it adapts you can teach it stuff and you can build automations with it so when there are certain tasks that are repeated you just tell atlas like do this when this other thing happens or do this when this trigger comes in or every month or something like that and it becomes kind of your assistant like your sidekick and uh you can build sub agents and set up atlas for specific tasks or specific sub projects so we have a few hundred of those that do um you know more narrowly scoped things um you know we have an agent that helps us do events like manage attendees and check people in um we have an agent agent that just does actually analysis on top of Atlas.

12:20It reports on how many users Atlas has every week, every month. So anyway, so we built all this stuff and we had talked to some other companies that were building it too and just said, well, why don't we try launching this and roll it out to people and see who else wants to use it? Because so many companies were having to build this themselves to get this experience. It's part of our labs teams. We have a team inside of WorkOS called WorkOS Labs, which is kind of building these new frontier experiences where we're not quite sure if we'll commercialize them or not. It's really the frontier of how we're growing and building as a company and how we're finding we can use AI software to drive forward our business further.

12:57I imagine some things will come out of that and turn into commercial products. Another one I can talk about in a bit is Airlock, which is our agentic permissioning system. It's a new thing for dynamic permissions for AI agents. That also came out of our labs team. but all this stuff is like you know compounding on top of the work we've done around authentication and identity and security and user management and sort of the enterprise like governance layer that we've built for many years and that you know that's being used by OpenAI and Anthropic and Cursor and SpaceX AI now actually separate with XAI and Perplexity and Sierra and you know Base 10 like all these companies Replit all these amazing companies so we've just kind of found we're in the middle position of all this stuff happening and it's very engineering led so the engineers are going to build they're going to build new things and and see what the future holds seems like you're pro engineering for life right not gonna stop that buck anytime soon just because agents do the coding doesn't mean we don't have engineers leading the agents right yeah i mean engineers at work os also do a lot of product stuff very small product management teams engineers are the ones shaping things, talking to users, driving them forward end to end.

14:10Our end customers are engineers. So it's like engineers building for engineers. We even find engineers do the marketing. It's something where it's very hard to do marketing if you're not technical for a technical product. I think we'll be quite engineering led forever as a business. Good. What it means though to be an engineer is changing. Like you said, now you're managing agents you kind of babysit multiple threads i think every engineer now is kind of like a manager sort of like exercising their executive function yeah to uh to operate and that's a harder job you can move faster you can build build new stuff we we we are structured so we have like very small teams that have like high agency and ownership over an area so they go really deep and own the thing end to end um and then just a bunch of those different teams you know working in different areas.

14:59And that's why we're able to have such a broad kind of product surface and yet be able to move so quickly. Yeah. A lot of things come to mind, a lot of technical things too, but I'm really curious as a, as a leader, how you keep up, like, do you, and you don't have to be, but feel free to be a little vulnerable. Like, do you feel overwhelmed? Do you feel it's hard to just maintain where things are going with so much going on, what the possibilities are how do you like from a person standpoint maintain you know when i was in college um was that uh was at at mit and there's this phrase that people say there it's like drinking from the fire hose and it feels like that right now yeah it does it's just it's just uh it's it's like this familiar feeling i have from my college days you know 15 years ago as a person how you feel about that though like drinking from the fire hose how does that do you feel like you enter every day with some version of balance or imbalance and i don't mean like just simply personally but it's like you you head to x.com which seems to be the epicenter again for everything which i'm kind of thankful because i just got tired of going everywhere i know not everybody is there but a lot of the action around ai and agents and this movement is really from my perspective and maybe years as well is on x i feel like every time i open up x i'm just like what's gonna happen today like not even just bad or good but like something big every day it really is like a day-by-day episodic thing what's going to launch this week what's happening next uh it can be exhausting just in that it's very unpredictable you don't know what's going to happen next but it's also really exciting i mean i you know i candidly i think sass got a little boring got a little stagnant you know we and she kind of and it wasn't that we weren't building interesting stuff or providing value for companies you know but it was this you know triple triple double double double you know hire the sales team turn the crank you know blah blah blah go public rule of 40 all this stuff we knew how to build the businesses like these businesses were kind of known the trajectory that they take and they were able to be financed in that way and and the kind of disruptive cycles were less frequent and usually more specialized to the domain of each SaaS product.

17:21All that's kind of thrown up in the air now because the agentic software is such a more compelling product experience for people. If you add this reasoning from the agents into a service, it blows away. It's like a magic ingredient. It blows away the previous stuff. Plus, you can build it faster. Plus, companies want to adopt the stuff sooner. you know it's this perfect storm of all these things together and so on the one hand i think it can be um upsetting i think for people that want or we're expecting the predictability of the market but that's not why i got into startups i mean i like the dynamic nature you know uh uh a friend of mine said once that uh startups are the extreme sports version of business absolutely and so i love it so i'm waking up every day you know going to sleep at one or two in the morning after you know vibe code building stuff at night and and waking up and can't wait to like scroll my x feed and be like who shipped what what are people talking about i love talking to people and arguing at x not not the way i'm trying to like get a rise out of people the disagreeing opinions i love it um the turning of an idea yeah i mean that's how we figure out the future is by discourse.

18:35It's by having conflicting viewpoints and discovering the truth through this kind of Socratic method. I spend a lot of my evenings, I live in San Francisco, a lot of my evenings going to meetups and talks and demo nights. And we host a bunch at our own office. We do our own big events. And some of that's for marketing for our products. But another reason is just for us to be part of the conversation and learn. So much of what I learned is from other people. It's not just from our team. And in moments of transition where everything is changing, it's the most interesting to me. I mean, that's what's exciting.

19:09So we have managed to find ourselves like right in the middle of it. It's pretty wild. This is not something I expected or anticipated that WorkOS would be a piece of infrastructure, like right in the middle of all the AI labs and the application layer companies and inference companies. Like if you go to our customers page we just launched a new one it's kind of like it's kind of like all the like we had a hard time picking what what logos to put on you know because because we have so many incredible businesses we get to work with um and i feel like i get to see a sliver into all of them i get to have be a very small part in all these businesses um and then help you know push forward what the future could be for software and tools i mean what a privilege to be able to do that it's it's a it's incredible yeah absolutely so so i'm having a blast i mean i'm sleeping sleeping less As my whoop says, my sleep isn't very good.

19:56But I am more excited about, you know, my job, what we can build as a builder, as a maker, our team, our culture, our customer base, and just what's happening in the technology industry. I'm more excited of the last six months than I have been my entire, you know, 15 plus career, 15 year plus career combined. This episode is brought to you by our friends at DeepGram. their speech models are built to understand words in context because context really matters, as you may know. Like if someone says there's a compatibility issue, they don't know if they're talking about technical difficulties due to conflicting interfaces or if they mean you want kids, I want a boat, this relationship is never going to work.

20:41Using context, Deepgram can tell one from the other, making conversations better. I use Deepgram. I love Deepgram. I use it for all my transcripts now. It is amazing. It's my intelligence packet and Deepgram's speech models and real-time infrastructure enables AI to keep up with the way people actually talk. Over 5 ,000 organizations globally, me included, use Deepgram's speech-to-text, text-to-speech and voice agent APIs in production. Built for accuracy, low latency, and cost efficiency, Deepgram makes high-quality voice experiments practical to build and operate at scale. Available through cloud APIs or self-hosted and on-premises deployments.

21:26Voice is the most natural interface, and Deepgram is bringing it to the technology we use every single day. Again, I'm using it. I love it. The future is talking, talking with Deepgram. And now when you spend$1 on Flux text-to-speech, you get$1 back, up to$500. To sign up for free or find out more, go to deepgram.com slash context is everything. Once again, deepgram.com slash context is everything. I'm trying to find the tweet because I thought it was pretty interesting that DHH of all people, as you may know, had said. He's so like that. I love it. You know, I was like, if there's anybody who's not, it might be David Hanemar Hansen.

22:18He went from zero to 100 real quick. He sure did. He was very, a year ago, he was very anti-AI. Yeah. he's like a bright slop you know it's bad you know it can't do this this this people that use it are dumb you know kind of like that now he was like i mean omar key's pretty incredible what he's done with it yeah exactly he says i love to see it i love to see it i mean people people should change their mind on this stuff they should be celebrated when they you know when they're they have that that plasticity um i think he's an inspiration to a lot of other people should be taking the same tack.

22:54Yeah, that's a loaded response in a way because some don't feel that way. I concur. I separate people from their politics and their provocateur nature. But let me read this tweet slash X post if you don't mind because it kind of sets the stage to some of the things you just mentioned with your whoop and your sleep and your vibe coding and whatever you want to call it and your 2 a.m. sleep times or whatever. He says, I know we have a lot of people and this is in regards to to omachi he says i know we have a lot of prs pending for omachi it was around 200 when quattro launched and now we're at approximately 1 000 that's not the point of the i'm trying to read here he says i promise they'll all get reviewed but i worked on quattro for three months straight and have been pulling 16 hour days since launch i don't know if david is a billionaire but he's probably pretty close.

23:47He's got a garage full of supercars. I mean, like that's whatever capitalism. I love it. But then he says, I might just need a quick break. He's talking in regards to like the momentum around Amachi. And then he says this, and this is the part I want to really kind of key into. And I want to see if you agree with this. He says, I work from home. So that helps in terms of like getting this time and doing the reps, but you can't do this pace forever. This is, and here's the part I want you to key in on. This is a seminal moment in computing history. This is a seminal moment. And so I'm up late, like you're up late.

24:22I don't run WorkOS. I haven't done what you've done, but I'm also very, very ambitious and doing cool stuff. But I feel like everybody is in that stage. So one, what do you think about this as a seminal moment? I'm sure you agree. But then two, zoom into what a day is like for you. Time you wake up, expose whatever you want or not want. What time you go to sleep. Just give me a glimpse into the crazy, maybe imbalance, or maybe struggle to balance life you're living now. Yeah. I mean, I don't think it's that out of balance, I guess. So I'll wake up around 7, 6.30, 7.30, something like that. And usually what I do is get up.

25:09I have this three-year-old golden retriever, so she'll sometimes wake me up on a letter outside. Um, make some coffee, uh, very quickly, like scan email and Slack for any kind of fires. Uh, but then usually I'm like trying to check in on what people are launching and reading X. Cause a lot of stuff does launch around then like for the day. Um, I usually spend about half an hour, like kind of working on my own stuff, catching up. And then usually I start working with the team around 8am, sometimes 730 around 8am. We have a bunch of people in the East coast also, so I can get started a little bit earlier.

25:42Yeah. I spend a lot of my time talking to our internal team, product team, engineering team. That carries me through early afternoon, usually. I spend a lot of time recruiting, talking to customers, partners. And so usually I'm pretty stacked. I will often work from home just because if I'm in our office, I'm in a conference room all day just on Zoom. Just back to back to back to back 30 minutes. And our customers, even if they're in SF, don't want to meet in person usually. They want to talk over Zoom. um so i'll do that and then uh depend on the day but i usually like to try to get outside for a bit the late afternoon in the bay area is always the best fog's burned off it's a little bit warm um so maybe go ride my bike somewhere or go to the gym you know between 3 30 to 6 4 to 5 is kind of my favorite and then most nights i'm off doing something i'm off like going to a dinner that i'm either hosting or another one that people are putting together we have demo nights at office we we order like pizza by the pound it seems like we have so much peas and soda um i host a monthly demo night where we have 20 developers come show off what they're building we do drink ups and we also just do a ton of stuff for the community like we um we do our big agent night mcp night events conferences um like last night i was at a thing watching uh talking with some people about uh well they were talking about neon and versell and how they interoperate and like new things that with the chat chupiti um uh kind of models and like how they're going to work with these agentic systems so i'm just like absorbing stuff all the time and meeting people and it's a really relatively small community in san francisco you see the same people again and again and again uh it's surprising um and then i'll get home i don't know anybody any time between like 9 and 11 if i'm out at dinner or staying late or grabbing a drink with someone and then usually pop over my laptop and you know crank on some stuff between then and I don't know midnight one two sometimes I see the clock at three and that's when I know I probably should go to sleep oh dang I know I hate when I'm like I look down on my head and I look down and I do work and I look back up I'm like oh my gosh this is an hour later than I thought it would be I really need to go to bed right yeah like right now it's like playing a video game you know or watching it reading a great book or watching a great tv show or something like one more episode um you know I I end up uh for my agents running them typically in fast mode because I don't have a big time interval where I'm like totally focused on it and I'll parallelize stuff and build stuff.

28:11But, um, so much of that is like me just trying to experience the frontier, what new things are getting shipped or launched. Um, and then on weekends, uh, it depends on the weekend, but I've had several where I just like kind of stay at home and grind through stuff and build stuff and read and just kind of learn and catch up and think, um, sometimes I'm out backpacking or going out. Like I really like cycling in the Bay area. So I've got a really long bike ride or something um but especially when it's raining like in the winter and just like burrow down you know make a cop pot of coffee on a saturday morning and just like you know ship 20 30 thousand lines of code that day that's a pretty that's a pretty good saturday yeah yeah yeah that's uh that's interesting you mentioned the word frontier and i think that's a very uh apt word to use to where we're at.

29:03David said, seminal moment. That is akin to this being a frontier moment we're all in. You got our friends over at AntCode. I feel like those guys are seriously on the frontier. They seem to be ahead of most things, but then also behind. I feel like you're ahead and you're behind, you're ahead and you're behind in any given moment. But when you look at the glimpse of the horizon, you personally have a vantage point on. and I'm speaking like to you specifically like what you get to see given the nightly events and the company you run and the things you're shipping and the internal teams and the labs and just the the inertia around what you get to experience what exactly how would you describe that frontier what exactly is that frontier how do you describe it it's a good question there's a lot of the stuff that I think you can articulate and then a lot of stuff that's more like gut sense intuition but high conviction that leads to investment in certain areas so like one thing that you see with developer especially developer focused companies is the kind of acceptance that agents will be the ones choosing our technical stack and kind of becoming the like consumer class to sell to you know when we launched this off.md thing it's really about allowing agents to go sign up for products and services like if you build an um you know an infrastructure service and you want to allow an agent just to go register without this the agent has to have an email address or somehow verify its identity somehow or it has to impersonate a human it can't show up and just say i'm an agent and get a you know api key and so we built a thing around this and it's kind of polarizing a lot of people are like that's crazy it's going to lead to abuse we don't need that it's fine for people to be in the loop but pretty much all the developer focused folks were like oh that's such a good idea because that's already happening they're already seeing the growth you know from that i'm with you on that yeah and so it's just not evenly distributed um i think right now the idea of agents being a big part of the workforce is like this um if you spend time on X, you'll see this with stuff like WorkOS Atlas, certainly the coding agents, you know, like people with Devon or Cloud Tag or things inside of Slack.

31:28And now with GrokBot, the GrokBot product. But I think that's still not common knowledge or accepted yet. Like most people in their jobs in the world are still just using ChatGPT on like cheap mode. They're not even doing longer term horizon reasoning or inference or tool calling or, you know automation or anything like that so it's uh it's very non-distributed yet like even if you go like we have an office in new york like i spend time there and it feels like new york is used to be six months behind now it's like a year behind where the bay area people are thinking and europe is like even further you know like several years kind of behind this this way of operating and thinking that's just that's just kind of the gut assessment i have that's not quantitative in any way.

32:17Anecdotal. This is from your perspective, remember? Yeah. Being on the frontier, I think, is that. It's like seeing what other organizations are doing, how they're trying stuff, how they're adopting. Something about working in the tech industry in the Bay Area is like... I have this joke that we actually all work for the same company. We all work for the same corporation, which is like just tech co, you know, and we all have different companies that are trying different projects and different things and trying to experiment with different ideas and different markets. But the Bay Area recycles teams and talents and process and tools and like open source is a big part of it.

32:59And, you know, the way the process of how people operate and work and, you know, what titles people have and how teams execute and stuff. All that's kind of the same. You know, it's kind of this industry of how we how we build software. And it's the reason why new companies can form and have like a culture already on day one it's because you're you're kind of working within the culture of the industry and uh there's a lot of sharing that goes on across people around ideas and tools it's it's much more collaborative than i think um people might expect it to be to be competitive even across the ai labs i think that's the big kind of the big dirty secret is like people bounce between these firms all the time you know people go from anthropic to open ai and then back to Anthropic or vice versa.

33:43Or, you know, they were at Anthropic and then they leave to go work at Thinking Machines or start another lab or they go to even DeepMind or other things. And so the techniques, the tools, the ways of building these things, the specific insights, it diffuses really quickly, you know, into the rest of the ecosystem. And I think that's generally good for the whole world. This episode is brought to you by our friends at Fingerprint. There's an interesting problem happening today, building products, you want to make sure things are as easy as possible for your users. But you also need to know who or what you can trust.

34:20That's getting more complicated today with AI. AI agents are browsing, they're shopping, they're transacting alongside humans. Some of that automation is legitimate. Obviously, I use AI, you use AI, but some of it is not legitimate. And fraudsters, the bad guys, are getting more sophisticated. That's where Fingerprint comes in. Fingerprint gives developers device intelligence to recognize returning visitors and make better trust decisions. It combines more than 100 signals across devices, networks, and behavior to identify visitors with precision, even as fraud tactics evolve. And it goes beyond identification.

35:01Fingerprint helps you detect bots, AI agents, proxies, VPNs, and other indicators of suspicious activity. So now you can protect accounts, prevent fraud, and reduce unnecessary friction for the people who actually use your product. And the developer experience is straightforward. One API call, no fraud detection model to build or to maintain. And now with AI install, you can add fingerprint to your platform directly from your favorite AI tool. So if you're building products for a world where humans and agents share the same Internet fingerprint is worth checking out. Head to fingerprint.com to get started.

35:41Once again, that's fingerprint.com.

35:48this frontier do you think it's i'm trying to pull out some of the threads that i think i heard you say the way we work is what i think is the frontier because when you say you know new york is behind a year and europe is behind three years and this is again from your perspective and you get to see this sort of inertia happening it's the acceptance of and the usage of agents, essentially, to do deep work, deep research, deep reasoning, and working autonomously for bounded time. Let's just say hours, maybe days, potentially. I know some folks are like, I got 17 agents working 24 hours a day, seven days a week.

36:31I mean, like, whatever, flex all you want. But the realness of that frontier, I feel like that frontier is accepting, because you kind of began with this identity factor, right? Even as humans, we have identity. Who are you? Who are your parents? Do you have brothers and sisters? Identity is formed in character, who you are, how you show up to the world, all those things, right? Then you hit this world where you've got an agent that mimics a lot of what we do as humans, thinking, reasoning, doing, working, et cetera. And then now we want to give them identities. I'm for that too. I think IDP for agents is a great thing because I want my agent to have some version of authentication and authorization and have guardrails and be bounded.

37:18This new word that seems to be everywhere now, bounded, and other many words that are clodisms or codexisms, whatever you want to call it. But I feel like it's this acceptance of an agent and working with an agent, some sort of autonomous unit, and we're on the precipice of that acceptance ad nauseum across the industry. It's almost by force. And then giving them a literal identity to say this is Bob at WorkOS. Maybe it's not really Bob at WorkOS. I'm just making that up. Maybe it's Atlas at WorkOS. You want to talk to that thing, that one thing and give it – I feel like that's the frontier. That frontier is accepting the agentic world and reorienting almost everything we do around that as a brand new OS and a brand new substrate.

38:09I feel like that's the frontier. And what comes from it is the innovation we've all been striving towards pretty much for the last 50 years in computing. I think that's well said. Yeah. The reason to embody it in something like Atlas is not because it makes the AI better. It's because it makes it easier for us to connect with. We as humans, we work with other people. We work on teams. We form relationships with others. We communicate with others to get stuff done. And so the ergonomics of collaboration are very human-centric. And in the same way we're building humanoid robotics, because we've sort of built the world for humans and we need to have robots interface with how the world is built, I think the same is true for digital AI teammates.

39:08It's, you know, they will talk to us over chat today. In the future, I think video and voice. Ooh. But it'll be because that's the way that we collaborate with each other. Yeah, we choose, yeah. It's like the APIs for human collaboration is, you know, text, voice, and video. And also documents and spreadsheets. Like an agent doesn't need to make like a spreadsheet. It can just store all the data in a database. It's not the representation it uses, but it will do those things to work with us to, to, as the kind of ergonomics for humans. Um, and so that evolution will be, you know, gradual in the same way, electric cars look like internal combustion cars, you know, and they have four wheels and a steering wheel and they drive and park in a garage and, you know, everything.

39:57And then over time, they'll start to look different as we discover the new kind of characteristics that can emerge from it. Like the new electric cars, you know, the self-driving ones um some of them don't have steering wheels the new zooks cars it just it's like kind of a box you sit in a different way is it really i was going to bring that up because waymos don't have the steering wheel right and that's but they're they're going to get rid of it you know if you look at the new waymos that they have the vans the steering wheel is much smaller it kind of looks like this kind of bolted on thing that you could just unplug it and throw it away um you know the steering wheel is a kind of this like backup right if if you go check out some of the self-driving cars they have in China.

40:37They even have these things with the steering wheel sort of fold into the console. It's almost like a debug port. It's like you pull out the steering wheel when you have to do it manually. It's the CLI of the car, man. It's like the terminal of the car, right? Yeah, it's dropped down to be low level. Look at the command line of this vehicle. Pretty much, yeah. It's like when the primary interface fails. What's this pedal on the right? Right. It's the gas, man. That's cool. Yeah, totally. Yeah. It's not gas. There's no gas anymore. It's like the save icon, you know, it's like what kids are like, what is that?

41:12What does that, what does that mean? Do you ever pause and ask yourself what's the point though? Do you ever do that? Do you ever pause and say, what's the point? Like, cause you mentioned a far off land called China, which to me, like I don't think about China a lot cause it's not that I'm not globally, you know, considered, but it's just like, it's just so far from me. It's not in my daily sphere to think about what the tech is somewhere else and how that's changing and there's people who travel and expose and there's people that have to work there because you're literally a global company do you ever think like what is the point of all this like even the self-driving cars like what is the point of all this innovation what is it the point of the ai stuff or of all of it you know like what's the point of i think it's better living through technology i mean that's like a that's that's a phrase that's as old as time.

42:04Is that right? Yeah. I mean, technology has always been the thing that's let us live longer, be healthier, connect better with people. I mean, it's also had a bunch of side effects and negative things. I think not intentionally. Social media, for example, has had a lot of negative issues, especially for kids. But it wasn't developed for that, you know clearly like it's not like the the folks at facebook were like trying to you know poison the brains of people or something it just it turned it turned into that as it ran up and went really quickly and i think that's that's the biggest concern with ai too is there's extraordinary power in these systems uh it's breathtaking like what they can do and and what they can make and what they can solve and and it's um it's going to have this extraordinary impact on the world in a way we can't even contextualize right now.

43:01Like we will look back on this in 10, 20, 30 years and the world I think will be almost irrecognizable. Yeah. In a good way, in a good way. Like looking back on like infant mortality, you know, being really high or like nutrition being really bad or, you know, like leaded gas, like there'll be all this stuff that will be like, I can't believe we dealt with those issues because we've just totally solved them. Right. You know, or solved it to the point where it's so rare that it occurs. These things that used to be just like common occurrences are now extremely rare that it's like a very exceptional.

43:37We just don't know what those things are, but inevitably there will be some places where we miss, where it has unintended consequences. I think that the question around alignment is really challenging. I mean, that's why it's so important to work on. But by and large, I'm pretty optimistic. I'm obviously very AI-pilled and working on this, But I think the benefit that it will bring to society and humanity and culture will far outweigh, far, far, far, far outweigh the negatives. And it's the job of leaders in the technology industry just to always have our eye on what could be the negatives so we can avoid the ones that we see in advance that we can.

44:13And then be really connected to people using these things. So when we do find areas where perhaps accidentally or perhaps just through lack of oversight, there's issues, we move to very quickly correct them. We don't let them fester. We solve them because those things have a tendency to run away and be very hard to solve later on if you don't solve them properly. So I think it's a big responsibility. It's a big privilege to be able to work in this way. But also, it's kind of like being a custodian of fire. Fire is really valuable. It's really useful. Like, you know, we can cook food because of fire.

44:48You know, our biology changed because of it. Like our ability to have homes and things changed. But it's also really destructive, you know, like it can run away. And harnessing it has been kind of a lot of what humanity is really, like tool builders and tool harnessers. And AI is like a version of that. I'm glad you said fire because that's the exact metaphor that I use when I talk to my wife. My wife is an avid chat GPT pro level user. She runs a custom tiny home company called Nomad Tiny Homes. And she is not by trade a home builder because that's what you have to be to build custom tiny homes is to be a home builder.

45:34and so she had to learn a lot of things through sheer force and literally not figuratively at all literally if there wasn't chat GPT to help her through all the churning and learning and responding and hiring and firing and customer comms and stuff like that, she would have been out to sea it's such a great thing you have It is the single biggest tool for learning. I think that's ever been built for education. For sure. Like it just, like if you can think it, you can build it. If you want to learn it, it's the most patient, deepest, smartest tutor you can ever imagine. Like if you just wake up one day and you're like, I want to learn calculus.

46:22I wasn't that good at doing math when I was in school. I kind of hated it. But you know what? I kind of want to learn it. Calculus seems kind of cool. And you could just be like the AI. Be like, yo, teach me calculus in a way. let's start from the basics and do it visually for me too so i can learn it build a little website for me to play around with make it a game i don't know like it's crazy what you can do with this stuff um it it's it's almost paralyzing like there's so many things to learn like i love learning things and reading books i had this moment the other day where i was like i have to pick what to learn because i could learn everything and i don't have time to learn everything oh yeah like there's so many things that are now accessible like do i do i want to learn how a jet engine works yeah kind of i mean it's pretty cool but like is it that useful or should i spend more time you know studying how like modern ai architectures work what's useful for my job what's useful for leadership yeah for nutrition for health you know for family like there's so many things you can do and so uh you know i i it's certainly the case as well that tech leaders have not done we have not done a good job like painting this rosy future um we've talked a lot about the risks and the um you know potential negative consequences of ai i i suspect this is in reaction to kind of what happened with social media and cambridge analytica and election interference and social media, all this stuff that the pendulum is kind of on the other side where we want to talk about the negatives so we get out in front of them.

47:59But every day, every month, to me it feels like the optimistic case in terms of what this will enable is more likely. Even if we stopped making the models better today and we just had what we have now for the next 10 years, it would be an amazing windfall for productivity across the world. And the models are getting better and better still. So it's bananas. That's the hardest thing for me. It's like, it's just kind of paralyzing waking up and just being like, oh man, this is, you know, it's like being on a rocket ship that's going to go light speed. And you're like, where are we going to end up?

48:36Does it matter or not? And you're such a small piece of it. You know, to me, I just, it's like, enjoy the process. like being part of it is such a cool thing like i'll like to be alive right now and to be part of this the acceleration to the singularity which is like what's happening in the futurist terms um just to be in that game you know to be an element in that uh what a wonderful thing privilege to be part of that this episode is brought to you by our friends at coder.com you know what your developers they have guardrails. The humans, guardrails, right? That makes sense. But your AI does not.

49:18How often are you running your AI dash dash dangerously? I think that most people are living a little dangerous. Some are sandboxed, some are not. Most are not. And I'm here with Nicky Pike, field CTO for Coder. So Nicky, what is the focus right now for Coder?

49:36MG: I mean, really our focus and what we're doing at Coder is on the governance around AI. So, you know, everyone right now, they're talking about, you know, AI agents like they're magic, but nobody's really asking where they run and what they have access to or what happens when one gets compromised. I mean, kind of give an analogy. You're not going to invite a contractor into your home and then leave and have them have full access to your safe and everything else on day one. You know, this is what people are doing with agents. We are wanting to, we are the infrastructure that helps you put guardrails and restrictions and governance around agents and what they're doing to you today.

50:10MG: That's where Coder succeeds because, you know, we started out as a CDE company. There's no secret. We were a CDE company. We created remote dev environments, standardized workspaces. But what we realized is that we accidentally built the exact foundation for the AI age before the AI age even really arrived. So isolated, governed, audible compute environments, that's what we build. And that's exactly what enterprises need when their workforce starts, including agents alongside their humans. So, you know, for us, this wasn't a pivot. This was really we finally found the problem that perfectly fit with what we'd already built.

50:45So if you're living dangerously, dash, dash, dangerously, non-sandboxed, you name it, Coder has AI governance for you, a centralized gateway to observe and to control LLM tool usage across your environments, authenticate model requests, audit every prompt, and contain agents behind a default deny allow list. The smart way. Learn more at Coder.com. Request a demo. oncemorecoder.com

51:22Yeah, you said fire. And I love the dovetail you just took us on because I totally agree with that. I feel anything I want to learn, I can learn it. Anybody with taste prior to the did great things, it's a multiplier of that original taste that you had. And you just kind of transplant that taste across things. But this idea that it's, like, for a while there, I was like, this is like the iPhone moment again. But that wasn't even enough. And I was like, this is like the internet moment. This is the moment the internet was invented. But that wasn't enough. It's closer to, like, both of those, plus cloud.

52:00And, yeah. And social networking. I think it's closer to fire. I think it really is closer to fire. It's a catalyst. It's a catalyst of change. Like, look how fast our industry changed, Michael. like our industry changed last year uh really you don't write your code anymore you're stupid oh my gosh i have 17 000 agents i write none of my code i'm cool about it you know or some are like not cool about it like the the shift i know we move fast in this industry but like wow that is not that's not adopting cicd that's a whole nother thing this is a whole nother thing it's very akin to fire and that's what i tell my wife as i was mentioning my wife and her company somebody babe this is it's like fire invented this this is like a new thing that nobody has ever harnessed or done and it's not like it's this it's a multiplier of what's already been there i watched i watched this i watched codex do this linux command today and i was just like wow when i saw it i was like cool you did that you solved my problem then i stopped and i'm like tell me how you did that yeah how did you do that how did you know that was the solution how did you know root cause how did you get to root cause because i want to learn enough to to direct or at least be a part of some of these things sometimes i'm like forget it man you're just so far ahead of where i'll ever be it's not a give up moment it's a it's more of a reverence almost in a way like this thing is so powerful to chain and pipe these commands that i would have never even learned about in linux and i'm like how did you do that i really feel like this is fire michael yeah 100 i mean it's uh and it's fun i mean that's that's the last thing for me it's like fun if all of this was just a big transformation and it was like powerful and you know valuable for the world and it was going to create a lot economic activity and and uh generate a lot of money and wealth and all that like that stuff's fine it's good um but it's also really fun like as a builder as someone that likes making things that's the reason i'm staying up late at night it's not it's not because i'm like worried about something i'm i'm not i'm not like you know twisting my hands and i i actually don't really feel stressed out i feel tired like it's it's fatiguing but it's more exciting than like any other time because there's so much possibility and like new opportunity next and so um i i stay up because there's like one more thing i want to make one more thing i want to build one more thing i want to try like the ideas every idea creates more new ideas you know on top of it um and that's such a cool thing uh i i hope that's something that everyone can experience i know people that aren't engineers or haven't built software maybe haven't experienced that because they don't have agency over their tools but it's it's the feeling i got when that made me want to make software like to do it professionally it's like you could make your own tools you can make your build your own legos you could play with these things and have ideas and then change your world you know have agency over it and that's uh i suspect that there's more people in the world than just software engineers many many many more people who have that curious creative spark inside of them and it just hasn't been you know there hasn't been the right canvas for them to do their work for that um and ai has the possibility to allow us to make a lot more of that and to be kind of like take the machines that we've had in these systems and and like make them more human to bring them to bring the computers closer to people rather than trying to bring people closer to the computers like we did all this stuff with the code boot camps you know like in the for sure 10 15 years and like that was about democratizing language coding yeah code.org you know it was like we got to take more people stem you know classes in college we got to get people closer to the computers to to to be productive and to get stuff done it was a big push now it's like for the first time ever we can get the computers closer to people and that's that's uh that's profound that's like completely a different a different change and wonderful english is the new language really i mean it's the new language of computers i think that's spot extremely spot on i like that take the thing i always try to share with people is like just do it it's just just try it that's our internal motto with ai you know it's like kind of the Nike just do it.

56:40It's like, you know, can it do this? I don't know. Could it do this? Like debate it. And it's just like, let's just try it. Let's just start, start the doing, you know? And, uh, and I always surprise myself. We always surprise, surprise each other. What's possible. Absolutely. Tell me if you know this person, this is a wise person on the internet. His name is Michael Greenwich. He said, uh, four hours ago, four hours. He's yeah. That guy, Four hours ago you tweeted this, posted this on X, however you want to describe it. And the reason why I want to end here or come close to the end here is I feel like we've talked about a lot of possibility.

57:16I want to kind of instill some hope if there isn't any or it's diminishing in our audience just because of the overwhelm in the morning. What should I do today? Oh, my gosh, I could do anything. What should I do? And is what I do actually valuable to the world? So I think you know what you wrote, but I'm going to read it in verbatim so the audience has this. you said starting a company is easier than ever than it's ever been. Ideas are easier. Shipping is easier. Fundraising is easier, but getting product market fit, which is essentially creating value, value that somebody exchanges value for, and that's monetary, right?

57:51But getting product market fit and creating a repeatable business is harder than ever before. Way, way harder. We've clearly been moved past the lean startup era. Now everything is extreme just to take a shot. Everything must be maxed out like you. You work on Fastboat. Tons of features, professional launch video, huge seed round, big parties, influencers, marketing stunts, and more. You've got to do it all just for a shot. And I agree with that. It's like the barrier to entry is lower but so high, but so high is what I read in there. The last thing to say is this seems unsustainable. I wonder how long can last and what comes after it.

58:38Sort of a big picture there, but the part I want to talk about is like, if we've got developers out there and audiences out there that maybe they haven't been developers, they've become developers because AI can teach you anything now and you learn, right? You can go from zero to one pretty quickly nowadays. This struggle to reach product market fit, where? Beyond just simply orchestrating agents. What is some of the value you get the glimpse? You mentioned slash customers from workOS.com, all the different folks you get to serve as part of your business. What exactly can folks focus on that isn't just simply this crucible of harnessing agents?

59:20I don't know. What is the product market fit? What is that nut there to crack? Yeah, there's a bunch of ways to go after finding product market fit.

59:34one way to do it is to go after something you've had a personal experience in, something that you had like an insight into something in the market, some pain. That one ends up becoming usually the place of the most durable information. Because if you're just trying to think your way in a PMF without having an experience yourself, it's very hard. It's essentially impossible, I think. so your personal experience you've had somewhere leading to pmf can can be helpful i think the other thing is just finding pain in the world um you know people talk about entrepreneurship is is a good good founders or entrepreneurs are like a heat seeking missile for pain and there's a long list of things you could build in the world that are nice to have that are kind of like the vitamin versus the painkiller but people don't really buy that many vitamins they don't take them every day they you know forget about them it's like a an optional thing but when when people have pain in their work or their life they immediately look to resolve it and it's really economically valuable for them so i think that's that's the thing to center into is like go find that in some way and then that doesn't tell you how to solve it you just find the problem and then your solution you just have to swing a bunch of times at it you know you can't think your way into the perfect solution we've pivoted and iterated a ton of times at work os even though the mission of the company has stayed the same i think the reason why is the pain point we were looking to solve never changed but the way that we've gone about solving it and the specific products and techniques have evolved a lot over the years um and then the last thing i tell people is uh before all of this you can pick your customers like you can choose what type of customers you want to serve and you should really like them because you're going to spend a bunch of time with them you know like if you really like doctors go build software for doctors you know if you don't like doctors don't go build software for doctors right don't do it like you can choose you know um if you really like building software for lawyers if you really like stuff for farmers or accountants or you know musicians or filmmakers or like whatever you know everybody in the world needs software and tools um pick the group of people that you really like not the person you want to be if you're like i really wish i was a musician so i'm gonna do software for you know music like you know you should probably just go become a musician then um but more so the people that you connect around if you're like uh you know you're you're kind of the tech guy hanging out around all the doctors well then your company should be the doctor software company and you'll have a great experience in your life making things for those people and it will feel really satisfying when you solve their problems because you'll feel this connection to them like i think that's that's that's the best part of the entrepreneurial journey is it's not making money it's actually like solving problems for people that you care about.

1:02:37So you can pick that when you get started, you can, you can center in that area and there's pain everywhere in the world. And anywhere you look, if you go deep enough, you kind of unravel it as like fractal. There'll be some thing that that person wants to be done better. Or as you know, if it was cheaper, it would change their business or their life. Or, um, there's a sticking point that they can't solve themselves. And then, uh, and then you just put your engineer hat on and you're like, all right, I'm going to go solve this. Yeah. This vantage point you have with working with folks and evening meetups and stuff like that, there are probably more than anything software engineers or newly founded software engineers.

1:03:18And what I heard you say there is pick a lane that you can kind of see some pain in. Have you seen any benefit to, let's say, a very capable software engineer aligning with a domain expert, right? If I can go pair up with an accountant, we can make a better thing for accountants. And it may not be QuickBooks, but it might be something below QuickBooks. But if I go pair up with a practitioner here locally in my town that's focused on health and wellness, and we both align, they're the expert in the health and wellness. I'm the expert in software engineering. We kind of pair up and we build them.

1:03:58Do you see that as sort of like a a new flywheel to start to create and spin? Yeah, definitely. Founding teams having domain experience and be able to connect with that audience so they can jump ahead to the solution.

1:04:18It's a really powerful formula with AI. It used to be in the past, you had to go find a technical co-founder and you wouldn't be able to scale without having someone like that. Not so much anymore. If you can think it, you can build it. You can at least build a version of it that's useful that you can use to validate it and get some initial momentum and traction. And then attract people to come build it and scale it with you. So I think the era of being non-technical is over. Saying I'm a non-technical co-founder is cope. It's like you can just go learn it. It's an indication of low agency. To some degree, it's always been that.

1:04:56But I think now it's that to a very large degree. because you can just go figure it out. Just try it. And it's not that hard. The amount of engineering skills you need to build software products, to build SaaS, you don't need to be the PhD computer science person working at Google for 15 years to do this stuff. The tools are good enough. The platforms are good enough. The systems are good enough that you can put it together. But you do have to have agency, and you do have to run through walls. If learning the code is going to be the thing that stops you from being successful, I hate to break it to you, but you're not going to succeed long term.

1:05:35Because that's not the hardest thing about building the business, is learning the code. That's one of the things you have to learn. But you've got to learn how to do sales. You've got to learn how to do marketing. You've got to learn how to do finance. Learning the code is just one of the things. And it's more learnable today than ever before. That's funny you mentioned that because one of the things I enjoyed most about, I think, our first podcast together, which was years and years ago now, like right around the middle, we started talking about sales. And I was like, let's stop talking about WorkOS and the technical details and auth and authentication, et cetera, because you were like sales.

1:06:10And I've been a tried and true salesperson my entire life. I grew up in sales. I don't think selling is manipulating. I don't think selling is convincing. I think selling is helping. I'm sure you agree because you do a great job having built your company. But that's exactly what you said so many years ago. And it's funny that you say that now. That's the critical thing to capitalize on is like software, mostly some version of easy, right? Some version of easy today. The hard part is interfacing with the world, serving them, selling them, fulfilling all the things. Always going to be hard. Always going to be hard.

1:06:49And there'll just be more need for it. We're entering this era where there's so much opportunity for software. The world, I believe, has an infinite appetite for software. We just can make software forever. And the constraint has always been people making it. And now that constraint is lifted. So I think the software industry is about to go through this explosion of usage and creativity. and we'll go from there being roughly 30 million professional software engineers in the world to at least 10x that in terms of people doing stuff, probably 100x in our lifetimes. Yeah, I mean, GitHub's hockey stick, I mean, GitHub's been the epicenter of our world for so long.

1:07:35If they're hockey sticking, is that even a word, a phrase? I don't know. If they're in the hockey stick moment, how does that multiply across the world? you know more if they go i think the the number was like commits per day was like half a billion 2023 to like 25 billion now i'm paraphrasing i there's like three charts i can't remember if i'm exactly saying but it was dramatic it was a dramatic uplift and that's get up like what does that do to mint brand new folks who've never written software never controlled software never wielded the creation of software good or bad to now go from well you have to be an engineer to well you don't but you need to know some things about engineering to do it well but now we've got this multiplication happening yeah but you gotta learn different things yeah it's like it's like you don't really have to learn like how to manage memory today like it's convenient if you know it yeah but you don't have to you know you can build so much stuff so much stuff without that um you know if there's 30 million software engineers pre-ai say we've 10x that that's like 300 million software people that can make software in the world wow you could kind of imagine that that'll happen like pretty fast you can sort of imagine like over the following 10 years plus 10 15 years that we'll 10x that again um so like 3 billion people that can like make a software thing you know chat with your phone and make a little mini app and share it with somebody you know that'll be like a thing people can do um and that's roughly half the population that's on the internet today.

1:09:10You know, it's about 6 billion and change people that are on online that use the internet in some way. And so it doesn't seem crazy to me to imagine in 10, 15 years that half of all people that are on the internet will be able to use something like lovable to make a service or like chat to PT. And I think that's, it's going to be just like the equivalent today of like using a phone. You know, it won't be the idea of making software. I think we still have this thing stuck in our head that it's like this hard to do thing. That's like, you know, it's, it's a very complex and requires a lot of skill and talent and all of that.

1:09:46It'll be as easy as speaking for people to do. And what a wonderful world that will be. I think what a wonderful world that would be. Yeah. Great place to end, Michael. Thank you so much for your time today. Thank you for coming back on the show. Always a fun time talking to you. Thank you for your wisdom you're sharing here on the, on the show, your anecdotes and hypotheses and all these things. I appreciate anything, anything left here in the closing. Anything you want to promote, share, announce, whatever. that Alice thing, if anyone wants to try it, come try it. Workquest.com slash Atlas. It's pretty cool.

1:10:18We'd love to hear feedback from people. And, you know, we're continuing to grow and hiring a lot and expanding. So I think for folks listening to this, especially engineers that want to work at that frontier, really want to be pushing things and building things in house. Yeah. We, we often hire people in who are kind of the most AI pill, people at their current job and they joined WorkOS and they're like, okay, now I'm amongst my people. So if you're listening to this and that's you, shoot me a DM. Yeah. I'm sad we didn't go deeper into Atlas. We sort of scratched some of the surface, but workOS.com slash Atlas.

1:10:53If you're on X slash Twitter X.com slash Greenwich is who you can follow to pay attention to most things from WorkOS. But again, Michael, thank you so much. Your Atlas is cool. Appreciate you. Thanks, Adam. See ya.

1:11:10Well, friends, what stays with me from this episode is MG's point that for the first time, we can bring computers closer to people instead of dragging people closer to computers. English. They speak English now. And if he's right, we go from 30 million people who can make software to 300 million or more, maybe 3 billion. The code gets easier. The hard parts don't. you still have to find real pain. Pick people you actually like serving and do the selling, which I've always believed is really just helping. And the pace we're working at has a cost, whether it's MG shipping code at two in the morning or DHH pulling 16 hour days.

1:11:50A few questions I want you to think about from this episode. If anyone can build a software, what makes yours worth choosing? Whose pain do you know well enough to solve? And when everyone has the canvas, what will you paint? Now, of course, I got some awesome sponsors making the show possible. Coder.com, ClickHouse.com, DeepGram.com, and Fingerprint.com. And of course, Fly.io. I love Coder. They're doing some amazing stuff around AI governance. ClickHouse, obviously, Observability, this new thing they launched. You can now, ClickHouse, way different than ever before. and Deepgram, they power all of my audio intelligence, all my transcripts, all things I do to make this show better, faster, easier, more possible and more amazing for you to listen to.

1:12:36And of course, our friends at Fingerprint, detecting what's fraud, who's real, who's not, amazing. Check them out, fingerprint.com. And of course, computers for agents from Sandbox to production, fly.io and my beat freak in residence, Breakmaster Cylinder. Bring those beats every single week. All right, friends, that's it. This show's done. We'll see you soon.

From the publisher

MG, founder and CEO of WorkOS, is back to talk about leading a company through what DHH called a seminal moment in computing. He explains Atlas, the AI coworker WorkOS built for itself and then launched, why agents will soon choose and sign up for your tech stack, and why this moment feels closer to fire than to the iPhone. Then he gets practical: starting a company is easier than ever, but product-market fit is harder, so be a heat-seeking missile for pain, pick customers you like, and stop calling yourself non-technical.

More from The Changelog: Software Development, Open Source

All episodes