AI transcript
0:00:07 Can you build an $8 billion startup by yourself using AI agents?
0:00:09 Well, today, we’re going to find out.
0:00:13 We have on Matan Grinberg, the founder of Factory.ai,
0:00:15 one of the best-kept secrets in Silicon Valley.
0:00:18 You know, everyone’s talking about vibe code this, vibe code that.
0:00:21 But as soon as you actually start vibe coding anything serious,
0:00:23 as of right now, it tends to break.
0:00:26 But with Factory.ai, you can actually build a real company
0:00:27 just using natural language.
0:00:30 Up until now, it’s only been used by huge companies.
0:00:34 But today, they’re releasing it to everyone and announcing it on this podcast.
0:00:37 So you guys are getting in on the ground floor.
0:00:38 So let’s just jump right in.
0:00:43 Cutting your sales cycle in half sounds pretty impossible.
0:00:47 But that’s exactly what Sandler Training did with HubSpot.
0:00:50 They used Breeze, HubSpot’s AI tools,
0:00:53 to tailor every customer interaction without losing their personal touch.
0:00:55 And the results were pretty incredible.
0:00:59 Click-through rates jumped 25%.
0:01:02 And get this, qualified leads quadrupled.
0:01:03 Who doesn’t want that?
0:01:07 People spent three times longer on their landing pages.
0:01:07 It’s incredible.
0:01:12 Go to HubSpot.com to see how Breeze can help your business grow.
0:01:17 Matan, man.
0:01:18 Thanks for coming on the show.
0:01:19 Thank you for having me.
0:01:20 It’s a pleasure to be here.
0:01:21 Yeah.
0:01:24 I’ve been thinking for a while, I really wanted to get you on here because I’ve been hearing from
0:01:28 friends in Silicon Valley for the last several months that, you know, basically factories like
0:01:30 Devin, but actually works.
0:01:31 At least that’s what they’ve been telling me.
0:01:33 Then I looked into your website, you know, and I was surprised.
0:01:36 I mean, I think you guys have been doing this for almost two years now.
0:01:40 You have incredible investors and you got Sean McGuire from Sequoia, one of the top investors
0:01:42 in the world who helped fund SpaceX.
0:01:44 Your background’s, you know, absolutely amazing.
0:01:48 You know, you’re a physicist who, my understanding is you published a paper with like the Einstein
0:01:49 of our generation.
0:01:53 And I looked at the website and to me, you know, your approach seems more practical.
0:01:56 Devin seemed to be kind of pitching, hey, we’re going to replace all of your engineers.
0:01:59 You seem to be more about like you’re empowering engineering teams.
0:02:02 My question is, you know, why don’t people know about you guys yet?
0:02:04 Yeah, great question.
0:02:06 And first of all, I know Silas, he’s great.
0:02:06 I’m a big fan of his.
0:02:09 But yeah, you know, there are a lot of different players in the space.
0:02:10 You know, we’ve been around for two years.
0:02:16 Our approach has been very much kind of disciplined in the sense that we’ve been building for enterprises
0:02:17 from day one.
0:02:23 And kind of top of mind, we knew that it’s a very tempting game, you know,
0:02:27 going into like X and LinkedIn and kind of playing that game.
0:02:32 And well, I think it’s really important to get out there and, you know, have developers
0:02:34 share their thoughts on both your vision and your product.
0:02:40 I think for us, we first wanted to really battle test our ideas and our product in the enterprise
0:02:46 where, you know, we can make bets that might not be, you know, that appealing for like a viral demo,
0:02:51 but might be very appealing for an enterprise or a developer, you know, working on some very nasty
0:02:53 cobalt migrations in like a 30 year old code base.
0:02:54 Right.
0:02:58 And so we’ve been very heads down for kind of the first basically two years of our existence,
0:03:04 just working with enterprises, deploying factory to these enterprises, improving the product from
0:03:04 there.
0:03:07 And so that’s why we’ve been a little less kind of outward.
0:03:10 But over the next couple of weeks, we’re at a point where it’s pretty important for us
0:03:14 to change that and, you know, be a lot more open, get a lot more developers in the platform,
0:03:19 because we’re at a point where we’ve found not only does the factory platform dramatically help
0:03:25 enterprise engineers, but we kind of naturally see them start bringing factor to like their side
0:03:28 projects or trying to sneak it into other avenues.
0:03:31 And yeah, it’s really at the point where, you know, we feel the product is mature enough to
0:03:34 start appealing to even more audiences than just the enterprise.
0:03:38 You know, so we’ve been very disciplined and very focused in our targeting for the first
0:03:39 two years.
0:03:43 But now we’re very excited to be opening up factory for GA access.
0:03:45 Yeah, that makes a lot of sense to me.
0:03:46 But maybe our listeners are like, what’s Devin?
0:03:50 Maybe like at a high level, we should explain to like, what is factory?
0:03:51 Like, what do you guys do?
0:03:52 And why should people care?
0:03:55 Yeah, so factory, we are building droids.
0:03:56 Okay, robots.
0:04:02 Yeah, it’s basically software robots that, you know, solve all the ugly tasks in the software
0:04:02 development lifecycle.
0:04:06 And so a lot of people, especially, you know, if they’re not that familiar with engineering,
0:04:08 might think engineering is just coding.
0:04:12 The reality is, especially at these large enterprises, developers don’t just spend their
0:04:16 time writing code, but they actually spend a majority of their time on all the stuff that
0:04:19 goes in before the code and all the stuff that goes in after.
0:04:22 And so that’s like understanding and planning and PRDs and design docs.
0:04:24 And then you get to the coding part.
0:04:26 And then after the code, there’s the review, the testing, the maintenance.
0:04:27 Yeah.
0:04:30 You had to go through like coding hell to get to like the actual fun part of coding.
0:04:31 Exactly.
0:04:33 And so we’re kind of focused on the whole like end to end.
0:04:37 And the reality is, I think it’s important to have that kind of holistic focus.
0:04:41 And in a similar sense to, you know, sometimes for self-driving, there are different approaches
0:04:43 about like which parts you need data from.
0:04:45 And I think our approach is very much end to end.
0:04:48 So my understanding is you guys are doing general release now.
0:04:51 So you’ve been serving enterprise customers, kind of silently building this for two years,
0:04:53 making great progress.
0:04:56 And a lot of head-to-heads, I’ve heard you guys even have been beating Devin often, which
0:04:56 is awesome.
0:05:00 And so now that it’s actually out there, maybe we can just like show people what it can do.
0:05:03 We could talk forever, but if they actually just see it, I think that’s going to speak
0:05:04 a thousand words, right?
0:05:04 Yeah.
0:05:05 Let’s jump into it.
0:05:06 So what are we going to do today?
0:05:09 So obviously, you know, the use cases that we do in the enterprise are, you know, obviously
0:05:12 very valuable, but a little less visually pleasing.
0:05:12 Right.
0:05:17 And so, you know, some common things that we do is like nasty migrations of like Java 7
0:05:23 to Java 21, database migrations, spring boot migrations, fun like SQL or COBOL.
0:05:27 Like all this stuff is very high value to the enterprise, but I think less maybe appealing
0:05:28 to the broad audience.
0:05:30 Yeah, I get like 50 views on YouTube or something.
0:05:31 Yeah, exactly.
0:05:33 So I was thinking, you know, for your audience, it might be fun.
0:05:37 People meme a lot about the number of engineers that work at DocuSign.
0:05:37 Oh, yeah.
0:05:40 It’s like 7,000 something crazy amount.
0:05:43 You know, every so often I’ll see a tweet that’s like after, you know, one company will announce
0:05:48 like a lot of layoffs, they’ll be like DocuSign announces hiring 100,000 more engineers
0:05:48 or something.
0:05:48 Yeah.
0:05:49 Yeah.
0:05:53 So I thought it could be fun to build a little toy version of DocuSign within Factory just
0:05:57 to get a sense of how Factory helps, you know, agentically automate some development
0:05:58 tasks.
0:05:58 Cool.
0:06:00 So we’re going to like build like a billion dollar company.
0:06:01 Yeah, let’s see what we can do.
0:06:03 Am I like a 50-50 partner in this or like how does this work?
0:06:06 Hey, if you help guide the droids, then you got to stay.
0:06:07 Okay, cool.
0:06:08 All right.
0:06:11 So here you can see we’ve landed on the Factory dashboard.
0:06:13 We have a chat interface.
0:06:15 So, you know, pretty standard, nothing too new here with LLMs.
0:06:17 We also have these four droids.
0:06:18 And so these are different.
0:06:19 These are our agents.
0:06:22 We call them droids because agent is kind of synonymous with poor quality.
0:06:23 Right.
0:06:25 It’s for, you know, different tasks that you might want to do.
0:06:28 So, you know, the knowledge droid is maybe you’re trying to create a design
0:06:31 dock for some large engineering tasks that your org will be doing.
0:06:33 Again, that’s a little bit more enterprise angle.
0:06:37 Code droid, this is really the kind of the one that everyone really wants to see, which
0:06:41 is, you know, whether it’s zero to one or end to end plus one, having the ability to go
0:06:44 into a code base and kind of build end to end features.
0:06:48 There’s the reliability drug, which can integrate with tools like Sentry or Datadog.
0:06:53 And if you have any outages, it can go in and create an RCA and even solve those issues.
0:06:56 We have the tutorial, which just guides you how to use Factory.
0:06:57 But cool.
0:07:00 I like the approach with the droids because it seems like you have these like specialized
0:07:05 droids that do a specific thing versus a lot of these, you know, AI agent companies.
0:07:09 I’m friends with Yohei Nakadima who did Baby AGI and actually like kind of spread that on
0:07:09 Twitter when it first came out.
0:07:11 And it was a great concept.
0:07:15 But a lot of these like that and Devin, they kind of promised they’re going to do everything
0:07:17 and then they kind of fail at almost everything.
0:07:21 I like the fact that you guys have like specialized ones that do specific things, have specific
0:07:22 deliverables.
0:07:23 Yeah, totally.
0:07:23 Yeah.
0:07:26 And I think, you know, the name of the game with agents is you want them to be as reliable
0:07:27 as possible.
0:07:32 And so if you could have them focused on certain core competencies in their workflows, it makes
0:07:34 that reliability a lot easier.
0:07:38 And so I just hit enter on, hey, let’s build out a toy version of DocuSign from scratch.
0:07:42 And so the code droid is now going in, creating this plan.
0:07:44 So, you know, here we have this general implementation plan.
0:07:49 Project setup, user authentication, document management, potential challenges.
0:07:53 Now, if I was going to be doing this a little more thoroughly and, you know, for production
0:07:56 grade right now, I might, you know, be a little more thoughtful in my responses.
0:07:58 But I defer to you.
0:08:04 Let’s just get a quick version up and running locally.
0:08:08 And then we can iterate from there.
0:08:13 So, you know, it’s asking me some very thorough questions like what tech stack, which core features
0:08:15 do you like this system to include authentication?
0:08:16 These are good questions.
0:08:17 I’m just saying, you know what?
0:08:18 You pick.
0:08:19 I don’t have time for this right now.
0:08:23 So, yeah, obviously in the enterprise setting, you’re going to be much more picky about this
0:08:25 when you’re not going, you know, purely from scratch.
0:08:29 But, you know, we can see it’s thinking process here as it’s going in and starting.
0:08:33 And so now it’s going to go and see here we have it running a command locally.
0:08:37 So the point here is that there are a lot of agents that run purely locally.
0:08:41 That’s like a tool like Cloud Code or any of the IDE agents.
0:08:42 They only run in your local environment.
0:08:47 And then there are some of the other agents like Codex or Devon, which only run remotely.
0:08:50 What’s incredible about Factory is we have the ability to do both.
0:08:52 And you can parallelize in both.
0:08:55 So as you’re delegating tasks, you can say some, you know what?
0:08:56 I’m confident in our tests.
0:08:58 I’m confident in our acceptance criterias.
0:09:00 I’m just going to go delegate that and send it to the cloud.
0:09:03 And then there are some where it’s like, I actually want to be pretty involved.
0:09:05 So I want this to work on it agentically, but locally.
0:09:09 Before you go on, I would love to like, so like, I believe I get what you said
0:09:10 in terms of being local and there’s a cloud.
0:09:13 And I think you said it, Greg Brockman was talking about this too, right?
0:09:16 That like that’s part of the vision in the future for OpenAI with Codex
0:09:17 is like that’s going to eventually do that.
0:09:20 And then you were telling me like basically Factory already does
0:09:22 what OpenAI is eventually doing.
0:09:22 Yeah, that’s right.
0:09:24 What’s the benefit of that kind of hybrid approach?
0:09:27 I mean, it allows these systems to have a silhouette
0:09:31 that’s much more similar to what we as humans have a silhouette.
0:09:32 So like, for example, when I’m working,
0:09:35 that looks like an agent working locally on my device.
0:09:37 Whereas when a colleague of mine is working,
0:09:41 it’s essentially equivalent to like some cloud environment,
0:09:42 you know, writing code.
0:09:44 And then how do I see the code that they write?
0:09:46 Well, they’ll submit a PR and then I’ll go and maybe look at their branch.
0:09:46 Right.
0:09:49 And so this way, it allows you to kind of spin up either
0:09:51 more copies of yourself or more copies of your colleagues.
0:09:52 Makes sense.
0:09:54 And so it’s kind of like, you know, a manager might think,
0:09:56 hey, here are these tasks that I’m going to do.
0:09:59 Here are these tasks that I’m going to kind of delegate to a colleague.
0:10:02 So you as kind of a pilot of this ship,
0:10:06 you kind of get to say, hey, which of these tasks will I monitor?
0:10:07 Which am I going to go send off?
0:10:09 And, you know, I’ll see the PRs later.
0:10:11 Yeah, so local is like giving you superpowers
0:10:13 or remote is like giving your team superpowers.
0:10:13 That’s right.
0:10:16 And with factory, you kind of get the best of both worlds.
0:10:16 It’s awesome.
0:10:17 Exactly right.
0:10:19 And so what happened here is, you know,
0:10:21 it was asking permission for certain commands.
0:10:23 I just turned on auto save and auto run.
0:10:26 So now it’s kind of just like full autonomous mode.
0:10:27 We’ll just like go run the commands.
0:10:30 So you can see it’s creating some repos, creating some folders.
0:10:33 And now it’s just up and running, you know,
0:10:35 and it’s going to create a few files
0:10:37 because you can’t make DocuSign just in one file.
0:10:40 So we can kind of have this running in the background here
0:10:41 and we can check in.
0:10:43 It should create all these files and then spin it up.
0:10:45 And so we’ll check it out.
0:10:47 Yeah, so this is basically like your YOLO mode or something?
0:10:48 Like it’s just like…
0:10:49 Basically, yes.
0:10:51 Except what’s nice is when you serve enterprise,
0:10:54 YOLO mode is not something that anyone ever wants.
0:10:55 Right, right, right.
0:10:56 We take this very seriously.
0:10:59 And so not only do you have this ability to just have like,
0:11:01 look, there are different levels of risk
0:11:03 for the auto accepting of CLI commands,
0:11:04 but also as an admin,
0:11:07 you’ll have the ability to whitelist or blacklist certain commands.
0:11:10 So you might not want to allow any pseudo commands
0:11:12 because it could do some pretty serious damage
0:11:14 or you might want to really restrict which folders
0:11:16 you’re even allowing the agent to get to.
0:11:18 How do you make that list of commands that can’t be accepted?
0:11:19 It’s in an admin setting.
0:11:20 Okay, cool.
0:11:21 Yeah, it’s just going to be running.
0:11:22 It might take a few minutes.
0:11:24 We can see it’s, you know, setting up its environment file.
0:11:26 But I think, you know, to your earlier question
0:11:28 about some of the things that I think OpenAI mentioned
0:11:29 in the codex launches,
0:11:32 it’s pretty clear that software development
0:11:35 is going to change dramatically over the next five years.
0:11:38 And I think an incongruence that currently exists
0:11:42 is that everyone says it’s going to change dramatically.
0:11:44 Yet the current paradigm,
0:11:49 pretty much what it looks like is putting AI onto existing workflows.
0:11:49 Right.
0:11:50 Right.
0:11:52 Like the existing workflow that developers have had for the last 15 years
0:11:54 is working in the IDE
0:11:56 and writing every single line of code there.
0:11:58 So we’ve applied AI into the existing workflows,
0:12:00 which is these AI IDEs.
0:12:02 But the reality is every big platform shift
0:12:05 has involved very significant behavior change.
0:12:07 You know, in the internet transition, what happened?
0:12:09 People went from getting most of their information from books
0:12:10 to like doing this,
0:12:11 and that’s how they get their information.
0:12:13 And mobile, what happened?
0:12:14 People went from like walking around with their heads up
0:12:16 to like walking around like that.
0:12:16 Yeah.
0:12:18 Very visceral behavior changes.
0:12:20 Yet AI, which is supposed to be the platform shift
0:12:22 that puts all these others to shame,
0:12:23 what are the most used products?
0:12:27 Well, it’s like ChatGPT, Perplexity, Copilot, Cursor.
0:12:28 Well, ChatGPT and Perplexity,
0:12:30 that’s just Google with better results.
0:12:31 Right.
0:12:31 Right.
0:12:32 It’s the same behavior.
0:12:33 Chat with Google, basically.
0:12:34 Exactly.
0:12:36 And then similarly with like Copilot and Cursor,
0:12:37 it’s essentially the same behavior,
0:12:39 which is like the IDE behavior.
0:12:40 Now you’re maybe pressing
0:12:42 some slightly different keys more often,
0:12:44 but it’s not like a viscerally changed behavior.
0:12:45 Right.
0:12:46 And that’s because we haven’t hit
0:12:48 that full transformation yet.
0:12:50 And what we’re really focused on with our droids
0:12:53 and with the ability to have them local and remote
0:12:55 is this is the new behavior that’s going to emerge
0:12:58 where you’re not writing every line of code.
0:13:00 the center of gravity of software development
0:13:05 will change from coding to instead understanding and planning
0:13:07 and then testing to make sure that these agents,
0:13:09 when they go and submit their PRs,
0:13:11 they did satisfy the constraints that you had in mind.
0:13:13 Yeah, that makes a lot of sense to me.
0:13:14 I like the guy, Primogen.
0:13:15 Do you know him on YouTube?
0:13:16 Yeah, of course.
0:13:17 Yeah, yeah.
0:13:17 He’s awesome.
0:13:19 And he’s slightly warming up to AI now,
0:13:21 but at first he was really hating on AI.
0:13:22 And it’s just because he just, you know,
0:13:25 he’s passionate about the art of just coding itself.
0:13:26 I’m lightly technical,
0:13:27 but I’m more of a business person, you know, investor.
0:13:29 And I think about it as like, you know,
0:13:30 when engineering, you know,
0:13:32 the point is to create things, right?
0:13:33 To solve problems.
0:13:34 Right.
0:13:35 And that’s what engineering really is about.
0:13:37 But even if we love it, you know,
0:13:38 that doesn’t mean that’s the best way
0:13:40 to do things in the future, right?
0:13:41 That’s why when I saw Factory’s website,
0:13:43 I was like, this feels more like the future.
0:13:44 100%.
0:13:46 And also to your point, it’s about building things.
0:13:46 Yeah.
0:13:47 So what does it mean to build something?
0:13:48 Well, you have an idea.
0:13:48 Yeah.
0:13:50 And that idea is defined by some constraints.
0:13:52 Let’s say your idea was Spotify.
0:13:53 It’s like, okay, well,
0:13:55 you have this social music sharing app.
0:13:57 Obviously there’s a business side of like
0:13:59 having all the agreements with the record labels,
0:14:01 but from the product itself,
0:14:01 it’s like, you know,
0:14:04 some low latency, high fidelity music sharing
0:14:06 that has social features, playlists, all that stuff.
0:14:08 Well, those are certain constraints
0:14:09 that you have in your head.
0:14:09 Right.
0:14:11 And what you need to do to actually build that
0:14:13 is you need to turn those constraints
0:14:16 into machine readable language,
0:14:17 which is, that’s why we have these,
0:14:18 you know, programming languages.
0:14:21 But what’s getting unlocked now
0:14:23 is that translation from you to the computer.
0:14:25 It used to be, you know,
0:14:27 you needed however many years of an education
0:14:28 and years of experience to actually learn
0:14:29 how to do that translation.
0:14:31 You get to hire people and like hope
0:14:32 that they actually did the work
0:14:33 and actually did what they said they were doing.
0:14:34 Maybe raise money.
0:14:34 Yeah.
0:14:35 Also, it’s like, that’s difficult.
0:14:37 Not a lot of people have access
0:14:38 to be able to hire that many people
0:14:39 or to raise the capital to do that.
0:14:40 Right.
0:14:40 Whereas now,
0:14:43 if you’re able to translate those constraints
0:14:44 that you have in your head,
0:14:47 you can kind of speak to a tool like Factory
0:14:49 and it will translate those constraints
0:14:50 into the software itself.
0:14:52 And so it lowers so many of those barriers
0:14:54 and kind of, again, refocuses like
0:14:55 what has made the best engineers
0:14:57 and the best product thinkers.
0:14:59 It’s not that they know every little detail
0:15:00 about every little language,
0:15:01 but they’re the best at thinking
0:15:02 about those constraints
0:15:05 and understanding what does my customer want
0:15:06 and how do I translate that
0:15:07 into these constraints?
0:15:07 Yeah, totally.
0:15:08 Thinking more about the customer
0:15:09 and spending more time on that
0:15:10 and the experience
0:15:11 and what problems you’re solving
0:15:13 versus dealing with bugs.
0:15:13 Yeah.
0:15:14 You know, I was in Silicon Valley
0:15:15 for 13 years.
0:15:16 I coded on and off.
0:15:17 I never like was super hardcore
0:15:18 into coding,
0:15:19 but a lot of my friends were
0:15:20 and I was just,
0:15:21 you know, I just felt like,
0:15:21 God, it’s, you know,
0:15:23 I just want to like solve problems
0:15:24 for people and create cool stuff,
0:15:24 right?
0:15:25 And tell people about it.
0:15:27 And like every time I started coding,
0:15:28 it was just like bugs
0:15:29 would just like drive me crazy.
0:15:30 I’m like, why am I like spending my life
0:15:31 dealing with these stupid bugs?
0:15:32 And then what other people,
0:15:34 you know, talk about like that.
0:15:34 It’s like the great thing
0:15:35 to be working on bugs.
0:15:36 Like, Jesus, it’s not.
0:15:37 You got one life.
0:15:38 Why would you spend all your life
0:15:39 solving bugs?
0:15:40 You want to build things?
0:15:41 Yeah, you’re right.
0:15:42 Right.
0:15:43 It’s like no one enjoys doing it.
0:15:44 And that’s the equivalent of that
0:15:45 in like the enterprise is like,
0:15:47 there’s so many menial,
0:15:48 tedious tasks
0:15:50 that enterprise engineers have to do.
0:15:50 Right.
0:15:51 It’s not why they got into engineering
0:15:52 in the first place.
0:15:52 Right.
0:15:53 And even something that you mentioned
0:15:55 that I think is a really interesting point to me
0:15:57 is you want to build something cool
0:15:57 or make something cool.
0:15:58 But the thing is,
0:16:00 you can’t just go to an LLM and say,
0:16:01 hey, make me something cool.
0:16:04 Because when you say cool in your head,
0:16:06 there are certain things that that means
0:16:07 that the LLMs don’t know.
0:16:10 And so this new era of software developers
0:16:12 who are the like best going to be,
0:16:14 the best people will be the ones that
0:16:15 when they think cool,
0:16:16 they know, like,
0:16:18 how do you elicit that from the model?
0:16:18 Right.
0:16:19 Because if I just said,
0:16:20 make me something cool,
0:16:22 probably I’m not going to be happy with it.
0:16:22 Right.
0:16:23 Because in my head,
0:16:24 when I thought cool,
0:16:25 I really meant something else.
0:16:26 Right.
0:16:26 Right.
0:16:26 Yeah.
0:16:27 Yeah.
0:16:28 And this is perfect for people like me
0:16:30 because I find like when I talk,
0:16:31 I’m a lot less articulate than I write.
0:16:33 I love to like sit around for an hour or two
0:16:34 and think something through
0:16:35 and write it out
0:16:36 and maybe very precise.
0:16:37 Yeah.
0:16:38 I mean, I used to be a physicist.
0:16:40 I spent a lot of my time with mathematicians
0:16:41 and there’s so many people
0:16:43 who felt more comfortable
0:16:45 explaining ideas with math
0:16:46 than with words
0:16:47 because sometimes like words
0:16:48 are very difficult
0:16:50 and similarly like engineers,
0:16:51 like sometimes it’s very difficult
0:16:53 to describe what you mean in words
0:16:54 as opposed to in like,
0:16:55 you know, programming language.
0:16:57 But understanding how to speak that
0:16:58 to the models
0:16:59 without writing out all the code yourself,
0:17:01 I think that’s going to be
0:17:02 a very powerful skill.
0:17:03 It’s kind of awesome
0:17:04 that we’re having this conversation right now
0:17:06 while like a billion dollar company
0:17:07 is being built in the background.
0:17:08 Yeah, exactly.
0:17:09 Yeah.
0:17:10 Hopefully it works.
0:17:12 I’m always nervous for like founders
0:17:13 when they do a live demo with AI
0:17:13 because, you know,
0:17:15 you never know what’s going to happen.
0:17:16 It looks like it’s doing
0:17:17 roughly the right thing.
0:17:18 It’s making some layouts,
0:17:18 some loading spinner,
0:17:20 login, register, dashboard,
0:17:21 uploading document,
0:17:22 creating signature.
0:17:22 That’s good.
0:17:23 You know,
0:17:24 got to have signatures
0:17:24 and DocuSign.
0:17:25 So yeah,
0:17:26 it seems like it’s on the right track.
0:17:30 Hey, we’ll be right back to the show.
0:17:30 But first,
0:17:31 I want to tell you about another podcast
0:17:33 I know you’re going to love.
0:17:35 It’s called Marketing Against the Grain.
0:17:36 It’s hosted by Kip Bodner
0:17:38 and Kieran Flanagan.
0:17:39 And it’s brought to you
0:17:40 by the HubSpot Podcast Network,
0:17:42 the audio destination
0:17:43 for business professionals.
0:17:44 If you want to know
0:17:45 what’s happening now in marketing,
0:17:47 especially how to use AI marketing,
0:17:49 this is the podcast for you.
0:17:50 Kip and Kieran
0:17:52 share their marketing expertise
0:17:53 unfiltered in the details,
0:17:54 the truth,
0:17:55 and like nobody else
0:17:56 will tell it to you.
0:17:58 They recently had a great episode
0:18:00 called Using ChatTBT 03
0:18:03 to Plan Our 2025 Marketing Campaign.
0:18:05 It was full of like actual insights
0:18:06 as well as just
0:18:08 things I had not thought of
0:18:11 about how to apply AI to marketing.
0:18:13 I highly suggest you check it out.
0:18:15 Listen to Marketing Against the Grain
0:18:16 wherever you get your podcasts.
0:18:20 I mean, it seems like the fact
0:18:21 that you were like tapping
0:18:23 into like the existing systems
0:18:23 inside of companies
0:18:24 where they already,
0:18:26 you know, use ticketing systems
0:18:26 and everything else,
0:18:27 like it seems like
0:18:29 a more natural fit for enterprises.
0:18:30 And I think it makes a lot of sense.
0:18:31 You know, we talk about this
0:18:32 in the podcast a lot.
0:18:33 Like I had Greg Eisenberg on
0:18:33 and we talked about like
0:18:34 what’s the future of business
0:18:35 and things like that.
0:18:37 And it feels like a lot of companies,
0:18:38 they’re starting to wake up to this,
0:18:40 but in the AI age,
0:18:41 it’s going to be more and more important
0:18:42 to keep reinventing yourself,
0:18:43 right, as a company.
0:18:44 And if you don’t,
0:18:45 a lot of companies
0:18:46 are going to be out of your business.
0:18:46 I mean, I think we’re going to see
0:18:49 lots of like multi-trillion dollar companies
0:18:50 that people are going to be shocked.
0:18:51 They thought trillion was a big deal.
0:18:52 It’s going to seem like
0:18:53 a small number in the future.
0:18:54 And we’re going to see
0:18:55 a lot of companies
0:18:56 that you thought were going to be around
0:18:57 that are no longer going to be around.
0:18:57 Yeah.
0:18:57 And it’s because
0:18:58 they didn’t reinvent themselves.
0:18:59 And, you know,
0:19:00 if I was a big company,
0:19:01 I would be looking at something like this
0:19:02 and thinking like,
0:19:03 this is something I can actually
0:19:04 get my teams to use.
0:19:05 It fits into our existing systems
0:19:07 and it allows us to
0:19:08 make our teams happier
0:19:09 to actually do work they want to do
0:19:10 and to try new things
0:19:11 because they need to be
0:19:12 trying new things right now.
0:19:13 Yeah.
0:19:14 You know, it is a little bit scary
0:19:15 to think about like,
0:19:16 oh, there are a lot of companies
0:19:16 that, you know,
0:19:17 if they don’t reinvent themselves,
0:19:18 there are certain things
0:19:19 that become obsolete.
0:19:20 On a micro scale,
0:19:21 it’s a little alarming
0:19:22 or a little concerning.
0:19:23 But I think on a macro scale,
0:19:24 it’s happening
0:19:25 because things are becoming
0:19:26 more efficient.
0:19:26 Right.
0:19:27 which I think is actually
0:19:28 a great story, right?
0:19:29 Like, I mean,
0:19:31 something that we talk about a lot,
0:19:32 even when we build factories,
0:19:33 the Henry Ford quote,
0:19:33 which is,
0:19:35 if you ask people what they want,
0:19:36 they would say faster horses.
0:19:38 And sometimes you need to kind of
0:19:39 look past what people are asking for
0:19:40 and build that automobile.
0:19:40 Right.
0:19:41 And something that,
0:19:42 you know, is a big change
0:19:44 is like in a world with horses,
0:19:45 the structure of the economy
0:19:46 is very different
0:19:47 than a world with cars.
0:19:47 Right.
0:19:48 But net net,
0:19:50 the world is much more efficient.
0:19:51 And so short term,
0:19:52 there are things that change,
0:19:53 but long term,
0:19:53 you know,
0:19:55 people who want to visit their families,
0:19:56 they’re able to do so faster.
0:19:56 Right.
0:19:57 If you have a medical emergency,
0:19:59 you can get to a hospital faster.
0:20:00 And so it’s like,
0:20:01 sometimes when you have these
0:20:03 step function changes in efficiency,
0:20:05 like the way the world has been built
0:20:07 will change a little.
0:20:07 Right.
0:20:07 But net,
0:20:09 it’s kind of a narrative of like
0:20:10 things becoming more and more efficient.
0:20:12 And I think that’s always a good thing.
0:20:13 And the creative people,
0:20:13 the smart people,
0:20:14 the great engineers,
0:20:15 the great product thinkers
0:20:17 will get to work on higher leverage,
0:20:18 more efficient problems,
0:20:20 which I think is a net good for the world.
0:20:20 Yeah.
0:20:22 And I agree with the smarter people
0:20:22 getting in from this a lot,
0:20:24 but I also think people who were
0:20:25 just like average
0:20:26 will also do quite well
0:20:28 because as long as you’re very persistent
0:20:29 and willing to grind and hustle,
0:20:30 like,
0:20:31 I think this is going to be amazing
0:20:32 for those people
0:20:32 because like maybe before
0:20:34 they weren’t the best coder
0:20:36 and they’d go talk to some engineer
0:20:36 and be like,
0:20:36 hey,
0:20:37 here’s my idea.
0:20:39 The person would be in the back of their head
0:20:39 and go,
0:20:39 okay,
0:20:40 cool idiot,
0:20:40 you know,
0:20:41 and like actually judging them
0:20:43 based on what they judged their IQ to be
0:20:43 or whatever.
0:20:44 Yeah.
0:20:45 but now with tools like this,
0:20:47 they can just go talk to the AI
0:20:47 and build the thing
0:20:49 and spread it out there into the world,
0:20:51 which wouldn’t have been possible before.
0:20:52 So that’s super exciting to me.
0:20:52 Totally.
0:20:53 And I think that’s also,
0:20:54 it kind of hits on something
0:20:55 that I found very compelling,
0:20:57 which is a lot of people are saying,
0:20:57 oh,
0:20:58 you know,
0:20:59 all these models are kind of
0:21:01 commodifying IQ or intelligence.
0:21:01 Yeah.
0:21:02 So that now like,
0:21:04 it doesn’t matter if you’re not that smart
0:21:06 because whatever the model at your fingertips,
0:21:07 you can have whatever intelligence you need.
0:21:09 And so then the question is,
0:21:09 okay,
0:21:09 well,
0:21:11 a prior theory is that your success
0:21:13 or your ability to do well
0:21:14 is kind of defined by,
0:21:14 you know,
0:21:16 IQ and how hard you can work
0:21:16 and that sort of thing.
0:21:18 And in this world where like,
0:21:19 this work is via API
0:21:21 or this intelligence is via API,
0:21:22 what is this new metric
0:21:24 that will determine like success
0:21:25 or not success?
0:21:27 Something that I think is kind of interesting
0:21:29 is the idea of agency
0:21:30 being the determining factor.
0:21:32 So even if you’re not the highest IQ,
0:21:34 if you have the will to go
0:21:35 and build things
0:21:37 as opposed to kind of being passive and lazy,
0:21:39 that’s going to be the determining factor.
0:21:40 And if you have that agency,
0:21:42 you can go and use all these models
0:21:44 that have that intelligence on demand for you
0:21:45 that might be experts
0:21:46 in all these niche fields
0:21:47 that you don’t have time
0:21:48 to become an expert in.
0:21:49 I find that kind of interesting.
0:21:50 Yeah,
0:21:50 I think that’s right.
0:21:52 I still think IQ is going to be a big advantage,
0:21:52 honestly.
0:21:54 I think this is going to amplify people
0:21:55 who are high IQ.
0:21:57 Maybe they’ll get ahead even further.
0:21:57 I mean,
0:21:58 a question is how closely
0:22:00 is agency tied to IQ,
0:22:01 which I think is an interesting question.
0:22:02 Right, yeah.
0:22:03 Yeah.
0:22:04 That’s a whole other conversation.
0:22:04 Yeah.
0:22:05 Cool.
0:22:06 So it’s still working.
0:22:07 It’s creating a lot of files, yeah.
0:22:08 Yeah, one thing I was thinking about
0:22:09 is one of our first guests
0:22:11 was Ervin Srinivas of Perplexity.
0:22:12 Oh, nice.
0:22:13 And when he first came on,
0:22:14 I was worried that, you know,
0:22:15 OpenAI was just going to eat them
0:22:16 like very quickly and destroy them.
0:22:16 Yeah.
0:22:17 Hasn’t happened.
0:22:18 They’ve done incredibly well.
0:22:20 And I feel like one of the benefits
0:22:21 they’ve had is the fact that,
0:22:22 you know,
0:22:23 every month there’s a different model
0:22:24 that becomes the best model.
0:22:27 And they’re able to tap into that
0:22:28 and take advantage of that
0:22:28 versus,
0:22:29 okay,
0:22:29 it’s just ChatsBT.
0:22:30 It’s just, you know,
0:22:31 Claude or whatever.
0:22:32 I was sitting there thinking like,
0:22:33 well,
0:22:34 Google’s got jewels now.
0:22:35 OpenAI has codex.
0:22:36 It feels like the fact
0:22:37 that you guys can tap into
0:22:38 whatever model is the best currently
0:22:40 is a huge advantage.
0:22:41 Totally.
0:22:42 All right.
0:22:44 So we look like we are up
0:22:45 and running here.
0:22:46 So I’m just going to ask
0:22:47 to set this guy up.
0:22:48 We’re still connected
0:22:49 to my local machine.
0:22:50 So looking good.
0:22:51 And we should be able to
0:22:53 check out our little DocuSign.
0:22:54 Let’s see how it looks.
0:22:55 Cool.
0:22:56 Fingers crossed.
0:22:58 All right.
0:23:00 So let’s check it out locally.
0:23:03 And we are good.
0:23:04 Oh my God.
0:23:06 This is our little DocuSign toy.
0:23:07 You got a landing page.
0:23:07 I mean,
0:23:08 does it do anything?
0:23:08 I mean,
0:23:09 that’s,
0:23:09 yeah,
0:23:10 we got a landing page.
0:23:10 The colors,
0:23:11 you know,
0:23:11 could do some work.
0:23:12 Very blue.
0:23:14 Not sure how I feel about this logo,
0:23:14 but you know what?
0:23:16 We asked for DocuSign.
0:23:16 Let’s see.
0:23:17 Sure,
0:23:18 DocuSign would be fine with that name.
0:23:19 DocuSign toy.
0:23:19 Why not?
0:23:20 DocuSign toy,
0:23:20 yeah.
0:23:20 All right.
0:23:22 So let’s try a little login.
0:23:22 So again,
0:23:24 we see we’re missing some of these logos,
0:23:25 but we’ll get our designer in here
0:23:26 to make it look cooler.
0:23:28 But I think the demo account
0:23:31 is user at example.com.
0:23:32 Password’s password.
0:23:35 And it looks like we’re in.
0:23:37 It seems pretty sweet.
0:23:38 Jesus.
0:23:38 Yeah.
0:23:39 So,
0:23:39 you know,
0:23:40 we can make some templates.
0:23:41 We can take a tour.
0:23:42 Let’s upload a template.
0:23:43 Okay.
0:23:45 Let’s throw in a PDF.
0:23:48 Let’s throw in a factory’s one pager.
0:23:48 Yeah.
0:23:51 and I’ll send it to myself.
0:23:53 Upload the Doc.
0:23:53 Okay.
0:23:54 Looks like a Doc.
0:23:55 Let’s add a signature.
0:23:56 There we go.
0:23:59 Add an initial field.
0:23:59 There we go.
0:24:01 Let’s send it for signature.
0:24:04 And we can see I now have one awaiting signature over here.
0:24:06 So,
0:24:06 yeah,
0:24:08 I don’t know about this layout necessarily,
0:24:08 but, you know,
0:24:09 we can adjust it.
0:24:09 Okay.
0:24:11 Send me the Docs to own half of this,
0:24:11 okay?
0:24:12 We’ll launch it.
0:24:13 All right.
0:24:14 So,
0:24:14 here we see,
0:24:15 so I have one to sign.
0:24:15 So,
0:24:16 let’s go and sign it.
0:24:18 Click to sign.
0:24:19 Can add my signature.
0:24:21 Click to initial.
0:24:22 Add the initials.
0:24:23 Seems pretty legit.
0:24:24 Complete signing.
0:24:25 There we go.
0:24:26 And now we have one completed.
0:24:27 Not too bad.
0:24:28 That is amazing.
0:24:29 Oh,
0:24:30 and we have a little activity log here as well.
0:24:31 That’s pretty sweet.
0:24:33 Let’s go back to the dashboard.
0:24:34 Okay.
0:24:35 That is incredible.
0:24:36 Okay.
0:24:36 So,
0:24:39 Factory is amazing at coding and creating a SaaS app.
0:24:41 It overuses blue.
0:24:42 There we go.
0:24:42 Hey,
0:24:43 you know,
0:24:43 we can make it red.
0:24:44 And that’s the thing is,
0:24:44 you know,
0:24:46 we can go back into Factory and adjust from there.
0:24:47 I think,
0:24:47 honestly,
0:24:51 the first thing that would just make this look a lot nicer is all these like placeholder icons.
0:24:52 Right.
0:24:53 Like having those,
0:24:54 that would make this look pretty sweet.
0:24:54 Yeah.
0:24:56 Here we even have templates already.
0:24:56 Yeah.
0:24:59 I think a big thing would be maybe adjusting some of the ordering here.
0:24:59 Like,
0:25:03 I think this is a little too much scroll when there’s nothing populated in there.
0:25:04 But honestly,
0:25:07 the only thing missing for this is like hooking it up to a database,
0:25:08 like to a proper backend.
0:25:08 Right.
0:25:12 And then all of the compliance stuff that I’m sure DocuSign actually has to deal with.
0:25:14 But you probably could talk to a factory and get a lot of that done.
0:25:16 Maybe not all of it at DocuSign level,
0:25:18 but like close enough.
0:25:18 Right.
0:25:19 For one shot in 15 minutes,
0:25:20 this is a very,
0:25:21 very solid start.
0:25:22 Yeah.
0:25:22 Yeah.
0:25:24 So like if you’re an entrepreneur watching this,
0:25:24 I mean,
0:25:25 in theory,
0:25:25 you know,
0:25:26 DocuSign is like,
0:25:28 I think they’re like an $8.6 billion company.
0:25:29 Right.
0:25:30 Yeah.
0:25:30 So in theory,
0:25:33 you could create something at that level as a,
0:25:34 even a one person team.
0:25:35 Which is just mind blowing.
0:25:36 Pretty exciting to see.
0:25:43 And I think it’s really also cool because it just means that barrier to creating that next $8 billion company is that much lower.
0:25:44 Right.
0:25:45 because now instead of the,
0:25:46 let’s say 500 engineers,
0:25:47 you might’ve needed to make that,
0:25:48 you know,
0:25:48 you know,
0:25:49 kidding.
0:25:49 But you know,
0:25:52 it just allows you to build these things out much faster.
0:25:53 And even within the enterprise,
0:25:54 in large orgs,
0:25:58 there are a lot of times where there’ll be teams of like 20 building out internal tools.
0:26:02 And it’ll take them just a huge amount of time when really the internal tool is a means to an end.
0:26:02 Right.
0:26:03 It allows you to get to that end faster.
0:26:04 Very cool.
0:26:10 You’ve been serving the enterprises for the last year or two and now it’s generally available.
0:26:11 Is that what’s happening right now?
0:26:11 That’s right.
0:26:11 Yeah.
0:26:12 We’re fully GA.
0:26:14 We have a team’s plan now.
0:26:16 That’s just to start $40 a month.
0:26:18 You can invite other teammates for an additional,
0:26:19 at least for now,
0:26:21 $10 for every additional user.
0:26:21 That’s all?
0:26:23 That’s crazy.
0:26:23 That’s all.
0:26:24 That is all.
0:26:24 Okay.
0:26:25 Yeah.
0:26:25 I mean,
0:26:28 I think the thing that we’re seeing is just a lot of small teams are really liking this.
0:26:29 And,
0:26:29 you know,
0:26:31 our focus still remains on the enterprise,
0:26:33 but if there’s something that delivers value to,
0:26:34 you know,
0:26:36 a demographic that’s slightly different than what we initially targeted,
0:26:39 initially there was a reason for the sake of focus to not kind of open that up.
0:26:41 But now we have the scalability.
0:26:43 We just want to put this in more people’s hands.
0:26:44 It’s still early.
0:26:45 I’m sure there are going to be things that,
0:26:46 you know,
0:26:47 people have a lot of feedback.
0:26:49 There might be things that they want to adjust and we’re very eager to hear.
0:26:50 But yeah,
0:26:52 I’m excited to put it in more people’s hands.
0:26:52 That’s awesome.
0:26:53 I see where it’s at today.
0:26:54 I think it has so much potential.
0:26:55 I mean,
0:26:56 where do you think factory is going to go in the future?
0:26:56 Like,
0:26:58 let’s say this is a super successful,
0:26:58 you know,
0:27:00 GA launch and everyone loves it.
0:27:02 And when people start talking about Devin,
0:27:05 they now say factory and you become part of like one of the category leaders.
0:27:07 Like where does factory go in the next like three to five years?
0:27:08 Yeah.
0:27:08 I mean,
0:27:14 I think the big thing in the near horizon is being this kind of unified platform for software development.
0:27:14 Right now,
0:27:17 a developer kind of lives a very fragmented life between GitHub,
0:27:18 between their IDE,
0:27:19 between Slack,
0:27:21 between Google drive,
0:27:22 between notion linear,
0:27:23 all these different tools.
0:27:29 And in a similar way to how a startup like Rippling kind of unified HR and IT into one place.
0:27:38 That’s what we want to do with software development because there’s just so much time spent crawling between all these different platforms and kind of pulling in all that information.
0:27:48 And that slows down that journey from idea to feature and factory and our droids are going to have access to all of these tools and they’ll meet you wherever you need.
0:27:54 So just like anything that’s like kind of most convenient is you need to do the least in the way of getting there.
0:27:55 And I think for us,
0:27:57 we want to meet developers where they are in these existing tools,
0:28:04 but then also provide them this nice new comfy home within factory where they can even start their projects as well.
0:28:09 So whether you have a very long Slack thread that’s an important conversation about product feature that you want to build out,
0:28:10 you can then tag factory.
0:28:16 It’s going to go and start creating a first pass, either like design doc or even a PR based on that.
0:28:21 Or if you have a backlog to tickets in your linear or JIRA, tag a droid and it’ll go and submit a PR to solve it.
0:28:22 That’s amazing.
0:28:25 So like, like even executives could be talking in Slack and like, you know,
0:28:28 they have an idea versus like going and bothering the engineering team.
0:28:32 They can just like have a first pass at it and see if it’s actually close to what they were imagining.
0:28:34 And then maybe, then maybe handing off the engineering team to take it to the next level.
0:28:35 A hundred percent.
0:28:39 I mean, something we’ve seen like that kind of naturally emerged that we weren’t expecting is
0:28:43 PMs at some of these enterprises we deployed to kind of got their hands on factory.
0:28:49 We didn’t initially plan on deploying to them and it kind of raised the bar for what is like a demo or a proof of concept internally.
0:28:57 Or also, you know, there’s so many times where PMs would need to ping front end engineers or full stack engineers to change copy or to add a page or this or that,
0:29:00 which like just slows down the engineering org so much.
0:29:01 They hate it.
0:29:04 It’s like, let me go in there and just change some stuff that you could just do if you knew what the hell you were doing.
0:29:05 Exactly.
0:29:05 Exactly.
0:29:07 And so now they don’t need to do that at all.
0:29:10 And they’re bragging like, hey, look, I just shipped some production ready PRs.
0:29:14 And if you’re an engineering leader and you’re worried like, oh no, my PMs are going to start submitting a lot of PRs.
0:29:17 What’s great about factories, we also adhere to your best practices.
0:29:26 So if you have pretty thorough docs in your org about like, here’s how we write tests, here’s how we ship features, here’s like our contributing guides, the droids will adhere to that.
0:29:32 And so if someone’s trying to ship a PR, it’ll actually go in and make the changes to make it adhere to whatever standards that you have.
0:29:33 That’s awesome.
0:29:39 And so it kind of keeps in check and makes sure you’re not just like introducing a lot of kind of vibe coded PRs in there.
0:29:39 Right.
0:29:42 But, you know, adhere to the enterprise standards, yeah.
0:29:45 Yeah, somewhat contributed to the vibe coding trend with Rally Brown.
0:29:46 I think it’s cool.
0:29:48 I mean, I think for like simple little apps, it’s cool.
0:29:51 But I think for anything complex, it starts to break down.
0:29:57 And I think it’s brilliant how you guys have started with enterprises and now kind of work to where regular people can use it as well.
0:30:04 I think also companies like OpenAI, I don’t see them anytime soon, like building something for the enterprise where they build everything out that the enterprise would need.
0:30:04 Right.
0:30:08 They’re going to start with more consumers because I mean, ChatsBT is a consumer app as of now.
0:30:08 Yeah.
0:30:10 So I love the strategy.
0:30:12 You keep saying droids, like why?
0:30:17 I mean, I know you kind of touched on it earlier, but like, are you not concerned about George Lucas or anything like that?
0:30:18 Yeah, no.
0:30:22 So we actually initially were incorporated as the San Francisco droid company.
0:30:23 Yeah.
0:30:23 Okay.
0:30:32 We were advised by our lawyers that Lucasfilm is very litigious and we decided to rename to the San Francisco AI Factory.
0:30:34 But we really loved the name.
0:30:36 Honestly, our customers really love the name too.
0:30:40 I can’t tell you how many times people are like, I speak with the droids or something like that.
0:30:47 And so now I think it’s more just, it’s going to be a sign of success when we get our first cease and desist from Lucasfilm.
0:30:48 Crazy side story.
0:30:53 So I don’t know George, but actually the reason I have lore.com is I was partnered with Barry Osborne, the producer of Lord of the Rings and the Matrix.
0:30:55 We were trying to make movies do it together.
0:30:55 Oh, no way.
0:30:57 And yeah, it was crazy for me.
0:31:03 Like I was involved in crypto pretty early on and like I sold my startup, not for like a huge amount, but you know, I was kind of in between projects.
0:31:07 My buddy introduced me to Barry and I was like, wait, you want to work with me on this stuff?
0:31:08 And like, let’s try to do it together.
0:31:09 We like really hit it off.
0:31:10 Damn.
0:31:15 You know, almost became like a somewhat like almost like a father figure to me, even though we’re like business partners, you know.
0:31:19 And he started getting me involved in meeting all these amazing people in Hollywood.
0:31:22 And I was out in Japan and he messages me.
0:31:26 He’s like, hey, do you want to go to a Skywalker ranch and possibly meet George?
0:31:26 That’s crazy.
0:31:29 I was like, yes, yes, I want to go.
0:31:33 And I literally went from Japan back to San Francisco.
0:31:35 You know, I don’t know anything about Hollywood.
0:31:35 I grew up in Alabama.
0:31:37 I bought this fancy jacket and everything.
0:31:39 I thought it was gonna be cool going to hang out out there.
0:31:42 I fly back from Japan to San Francisco.
0:31:46 And then that was when the wildfires happened and the entire thing got canceled.
0:31:46 Oh, no.
0:31:47 The entire thing got canceled.
0:31:50 And so it just never happened again.
0:31:51 And then COVID happened right after that.
0:31:54 And so I never got to meet those people, which was really sweet.
0:31:55 Oh, man.
0:31:55 That would have been awesome.
0:31:59 So something I think, you know, I just mentioned my son, because actually he went back to San
0:32:00 Francisco with me.
0:32:04 One thing I think about a lot is I think about, you know, what should I be teaching my son?
0:32:07 In the AI age, should he be learning to code?
0:32:09 It’s hard to know what he’s going to do in the future, right?
0:32:09 Yeah.
0:32:11 You’re too young to have a kid right now, or probably are.
0:32:13 If you did have a kid, would you be teaching them to code?
0:32:14 Yeah.
0:32:16 I mean, I think 100% unequivocally.
0:32:23 Just like I think, you know, even though I don’t need to do multiplication that often these
0:32:27 days, I think understanding the things that underlie all of the technology around us will
0:32:28 always be important.
0:32:28 Right.
0:32:33 I think similarly, like understanding like machine code doesn’t really matter or like assembly
0:32:34 doesn’t really matter.
0:32:39 But to have that full kind of systems understanding of the different layers of abstraction will always
0:32:39 be important.
0:32:39 Right.
0:32:44 Whether you’re a software engineer, a product builder, whether you’re a theoretical physicist,
0:32:48 you’re like, it’s still important to understand kind of the bare bones of what underlies
0:32:50 whatever it is that you’re working on.
0:32:55 And I think what we’re coming to terms with is there’s kind of going to be a mountain of
0:33:00 material that we no longer need to know, but it’s still like you will be at a huge advantage
0:33:01 if you are familiar with that.
0:33:02 Yeah.
0:33:03 So I think coding is incredibly important.
0:33:07 I think mostly for the way it teaches you how to reason and how to think.
0:33:07 Right.
0:33:08 That’s what I was going to say.
0:33:10 In that systems way, thinking about constraints.
0:33:14 Again, I think that methodical way of thinking and reasoning through problems, that’s always going
0:33:15 to be valuable.
0:33:20 I was a physicist before, and there’s a funny thing where there are a lot of physicists in
0:33:22 a lot of the foundation model labs.
0:33:25 And is it because there’s a lot of black holes involved in the LRMs?
0:33:25 No.
0:33:26 Not as far as we know.
0:33:27 Yeah.
0:33:27 As far as we know.
0:33:28 I don’t know.
0:33:28 Yeah.
0:33:29 Maybe this is maybe this laid poorly.
0:33:35 But the reality is working on problems that have very difficult reasoning and require the
0:33:39 synthesis of a lot of different information and reasoning about it in a pretty like non-trivial
0:33:44 quantitative way and kind of a systems way, that is just a valuable skill no matter what
0:33:45 and whatever domain you end up applying it to.
0:33:46 Right.
0:33:48 And so I think this applies similarly with these tools.
0:33:52 Now, I think this next generation should not necessarily be brought up the same way we
0:33:55 were because they should also be native in how to use these tools.
0:34:01 Just like how in eras before, people would spend a lot of time with an abacus doing calculations.
0:34:04 It’s important to know, OK, I can do the calculation myself.
0:34:07 But once you’re past that, it’s like, OK, now use a calculator from now on.
0:34:12 Just remember, you still know how to do the math because now you can use it and have much
0:34:14 higher leverage to kind of build things.
0:34:15 I agree.
0:34:18 You know, I’ve seen a lot of people, at least on Twitter from Silicon Valley, like commenting
0:34:22 on this and they seem to have like really, I don’t know, they treat it too binary in my
0:34:22 opinion.
0:34:27 Like I’ve seen either like, yeah, my kid’s like seven and they’re like going through like a
0:34:28 coding boot camp.
0:34:31 Or it’s like, I’m never teaching my kid to code because it doesn’t matter anymore.
0:34:34 Neither one of these makes sense to me right now.
0:34:37 Like with my son, I’ve been like kind of easing him into things like I showed him the command
0:34:37 line.
0:34:38 He’s like, oh, that’s cool.
0:34:39 That’s how that works.
0:34:41 I’m like, yeah, this is actually what’s going on behind the scenes.
0:34:42 Yeah.
0:34:44 I taught him like four or five commands.
0:34:46 I was like, if he wants to play with it sometime, he can.
0:34:48 But I’m not going to like make him do that.
0:34:51 And then we played with Replit and a few other tools.
0:34:54 And he just like loved the idea of like making stuff with AI.
0:34:55 He thought that was fascinating.
0:34:56 So I’m trying to do both.
0:34:58 I think most people should probably be doing that.
0:34:59 Totally.
0:35:01 We got to get your son on factory.
0:35:01 Yeah, I will.
0:35:01 I will.
0:35:02 Be careful.
0:35:04 He may like end up taking over your company one day.
0:35:05 Yeah.
0:35:06 Maybe I’d like a fun question before we go up here.
0:35:10 You know, so Matan, imagine you have a time machine, you know, you’re a physicist.
0:35:13 If I’m saying anything stupid here, just, you know, don’t.
0:35:16 You have a time machine, you travel to the year 2050.
0:35:21 Let’s say you get out in San Francisco, you know, and what do you see?
0:35:25 What’s different in the physical world, digital world, life, whatever.
0:35:26 Yeah.
0:35:30 I maybe have a hot take here, which is I think the arc of technology is actually exactly an
0:35:33 arc in that kind of where did humanity come from?
0:35:37 But like in nature, like hunter gatherer, there was no discernible technology other than
0:35:39 maybe some sticks and stones and tools and whatever.
0:35:44 I think we’re kind of about to hit the apex where like now you look out in San Francisco
0:35:46 and you see like Waymo’s, you see so much technology everywhere.
0:35:50 I think as time goes on, we’re going to kind of go back down and reduce the presence of
0:35:53 technology as much as possible, which you can kind of see the early starts up.
0:35:56 Are you saying everyone’s going to like destroy all the Waymo’s or what?
0:36:01 No, no, no, but just like people, as we get so much like efficiency and so much value
0:36:05 out of like medicine and technology and all this, I think you’re already seeing like
0:36:06 early inclinations of people trying.
0:36:12 there’s all these movements about, you know, cold showers or like MMA and all, or all this
0:36:14 like, you know, people spending a lot of time in nature or getting all this sunlight and kind
0:36:17 of going against the like pure technology for everything.
0:36:18 So the tech’s going to be in the background.
0:36:21 It’s going to be, it makes things amazing and it’s not going to be in our face.
0:36:21 Yeah.
0:36:25 And now I think this is very much SF is a bubble and the rest of the world is going to look
0:36:27 very different and we’ll adopt things in a slower pace.
0:36:31 I think San Francisco is going to kind of come back around and have a little bit of a
0:36:36 like, how can we have the same enablement with minimal presence from these things?
0:36:42 This also is skipping over the probably 20 years of like robotics that is going to be
0:36:47 at that apex, which I think maybe at some interim between where we are now in 2050, we’re going
0:36:52 to have by far a larger robot population in San Francisco than human.
0:36:53 I agree.
0:36:53 And I’m here in Kyoto.
0:36:58 I’ve actually been, you know, trying to advocate for the US and Japan to work together on this
0:36:58 stuff.
0:37:01 Because I think Japan would be a perfect place to be testing the robots too.
0:37:02 People are super open to it.
0:37:03 Oh, a hundred percent.
0:37:04 Yeah.
0:37:07 I mean, growing up, I was inspired by Gundam and all that, which, which came from.
0:37:09 I almost wore a Gundam shirt today because the whole droids thing.
0:37:10 That’s awesome.
0:37:11 Yeah, I almost did.
0:37:12 I agree with that.
0:37:15 And actually, I think, you know, OpenAI just announced the whole thing with Johnny Ive,
0:37:15 right?
0:37:16 Where they’re going to be building devices.
0:37:20 I was like, in one of the first episodes of the next wave, I said that I was like, I think
0:37:23 one of the big things in the future was like a future prediction is that the iPhone
0:37:25 is not going to be the last device for humanity.
0:37:27 It’s not going to be the last way that we interact with technology.
0:37:31 And the fact that we’re all just like staring down at our phones right now, like, you know,
0:37:32 looking like morons.
0:37:35 I’m hoping that eventually goes away and we have better ways to interact with technology.
0:37:37 And I think AI will actually enable that.
0:37:39 So it kind of fits with your vision for the future.
0:37:39 Yeah.
0:37:40 This has been awesome.
0:37:42 Is there anything you want to tell people?
0:37:43 Like, how can they get started with Factory today?
0:37:46 Yeah, go to factory.ai and get started.
0:37:48 We have 14-day free trials for everyone.
0:37:50 So go in, check it out.
0:37:54 If anything’s not up to your liking, or if you have any questions or thoughts, just shoot
0:37:55 me an email, matonatfactory.ai.
0:37:57 Happy to jump in.
0:37:57 Cool.
0:37:59 Is there anything special for Next Wave listeners?
0:38:00 Yeah.
0:38:04 So we have a very special deal for Next Wave listeners.
0:38:05 Awesome.
0:38:07 I believe it should just be in the link in the show notes.
0:38:08 Awesome.
0:38:08 That’s great.
0:38:11 Should people follow you on social media or maybe like follow Factory?
0:38:13 I think it’s Factory AI on Twitter.
0:38:13 Yeah.
0:38:15 Factory AI on Twitter.
0:38:16 Give us a follow.
0:38:18 We’ll be posting all of our updates, demos, that sort of thing.
0:38:19 Well, that’s awesome.
0:38:23 For anyone listening, you know, we’re trying to level up our game with this podcast and
0:38:25 hopefully, you know, you’re finding the episodes useful.
0:38:30 So please, if you would, it would mean a lot to me if you would subscribe on YouTube, you
0:38:33 know, if you’re listening on Apple or Spotify, subscribe there.
0:38:34 And yeah, thanks, man.
0:38:35 It’s been awesome to talk.
0:38:35 Thank you.
0:00:09 Well, today, we’re going to find out.
0:00:13 We have on Matan Grinberg, the founder of Factory.ai,
0:00:15 one of the best-kept secrets in Silicon Valley.
0:00:18 You know, everyone’s talking about vibe code this, vibe code that.
0:00:21 But as soon as you actually start vibe coding anything serious,
0:00:23 as of right now, it tends to break.
0:00:26 But with Factory.ai, you can actually build a real company
0:00:27 just using natural language.
0:00:30 Up until now, it’s only been used by huge companies.
0:00:34 But today, they’re releasing it to everyone and announcing it on this podcast.
0:00:37 So you guys are getting in on the ground floor.
0:00:38 So let’s just jump right in.
0:00:43 Cutting your sales cycle in half sounds pretty impossible.
0:00:47 But that’s exactly what Sandler Training did with HubSpot.
0:00:50 They used Breeze, HubSpot’s AI tools,
0:00:53 to tailor every customer interaction without losing their personal touch.
0:00:55 And the results were pretty incredible.
0:00:59 Click-through rates jumped 25%.
0:01:02 And get this, qualified leads quadrupled.
0:01:03 Who doesn’t want that?
0:01:07 People spent three times longer on their landing pages.
0:01:07 It’s incredible.
0:01:12 Go to HubSpot.com to see how Breeze can help your business grow.
0:01:17 Matan, man.
0:01:18 Thanks for coming on the show.
0:01:19 Thank you for having me.
0:01:20 It’s a pleasure to be here.
0:01:21 Yeah.
0:01:24 I’ve been thinking for a while, I really wanted to get you on here because I’ve been hearing from
0:01:28 friends in Silicon Valley for the last several months that, you know, basically factories like
0:01:30 Devin, but actually works.
0:01:31 At least that’s what they’ve been telling me.
0:01:33 Then I looked into your website, you know, and I was surprised.
0:01:36 I mean, I think you guys have been doing this for almost two years now.
0:01:40 You have incredible investors and you got Sean McGuire from Sequoia, one of the top investors
0:01:42 in the world who helped fund SpaceX.
0:01:44 Your background’s, you know, absolutely amazing.
0:01:48 You know, you’re a physicist who, my understanding is you published a paper with like the Einstein
0:01:49 of our generation.
0:01:53 And I looked at the website and to me, you know, your approach seems more practical.
0:01:56 Devin seemed to be kind of pitching, hey, we’re going to replace all of your engineers.
0:01:59 You seem to be more about like you’re empowering engineering teams.
0:02:02 My question is, you know, why don’t people know about you guys yet?
0:02:04 Yeah, great question.
0:02:06 And first of all, I know Silas, he’s great.
0:02:06 I’m a big fan of his.
0:02:09 But yeah, you know, there are a lot of different players in the space.
0:02:10 You know, we’ve been around for two years.
0:02:16 Our approach has been very much kind of disciplined in the sense that we’ve been building for enterprises
0:02:17 from day one.
0:02:23 And kind of top of mind, we knew that it’s a very tempting game, you know,
0:02:27 going into like X and LinkedIn and kind of playing that game.
0:02:32 And well, I think it’s really important to get out there and, you know, have developers
0:02:34 share their thoughts on both your vision and your product.
0:02:40 I think for us, we first wanted to really battle test our ideas and our product in the enterprise
0:02:46 where, you know, we can make bets that might not be, you know, that appealing for like a viral demo,
0:02:51 but might be very appealing for an enterprise or a developer, you know, working on some very nasty
0:02:53 cobalt migrations in like a 30 year old code base.
0:02:54 Right.
0:02:58 And so we’ve been very heads down for kind of the first basically two years of our existence,
0:03:04 just working with enterprises, deploying factory to these enterprises, improving the product from
0:03:04 there.
0:03:07 And so that’s why we’ve been a little less kind of outward.
0:03:10 But over the next couple of weeks, we’re at a point where it’s pretty important for us
0:03:14 to change that and, you know, be a lot more open, get a lot more developers in the platform,
0:03:19 because we’re at a point where we’ve found not only does the factory platform dramatically help
0:03:25 enterprise engineers, but we kind of naturally see them start bringing factor to like their side
0:03:28 projects or trying to sneak it into other avenues.
0:03:31 And yeah, it’s really at the point where, you know, we feel the product is mature enough to
0:03:34 start appealing to even more audiences than just the enterprise.
0:03:38 You know, so we’ve been very disciplined and very focused in our targeting for the first
0:03:39 two years.
0:03:43 But now we’re very excited to be opening up factory for GA access.
0:03:45 Yeah, that makes a lot of sense to me.
0:03:46 But maybe our listeners are like, what’s Devin?
0:03:50 Maybe like at a high level, we should explain to like, what is factory?
0:03:51 Like, what do you guys do?
0:03:52 And why should people care?
0:03:55 Yeah, so factory, we are building droids.
0:03:56 Okay, robots.
0:04:02 Yeah, it’s basically software robots that, you know, solve all the ugly tasks in the software
0:04:02 development lifecycle.
0:04:06 And so a lot of people, especially, you know, if they’re not that familiar with engineering,
0:04:08 might think engineering is just coding.
0:04:12 The reality is, especially at these large enterprises, developers don’t just spend their
0:04:16 time writing code, but they actually spend a majority of their time on all the stuff that
0:04:19 goes in before the code and all the stuff that goes in after.
0:04:22 And so that’s like understanding and planning and PRDs and design docs.
0:04:24 And then you get to the coding part.
0:04:26 And then after the code, there’s the review, the testing, the maintenance.
0:04:27 Yeah.
0:04:30 You had to go through like coding hell to get to like the actual fun part of coding.
0:04:31 Exactly.
0:04:33 And so we’re kind of focused on the whole like end to end.
0:04:37 And the reality is, I think it’s important to have that kind of holistic focus.
0:04:41 And in a similar sense to, you know, sometimes for self-driving, there are different approaches
0:04:43 about like which parts you need data from.
0:04:45 And I think our approach is very much end to end.
0:04:48 So my understanding is you guys are doing general release now.
0:04:51 So you’ve been serving enterprise customers, kind of silently building this for two years,
0:04:53 making great progress.
0:04:56 And a lot of head-to-heads, I’ve heard you guys even have been beating Devin often, which
0:04:56 is awesome.
0:05:00 And so now that it’s actually out there, maybe we can just like show people what it can do.
0:05:03 We could talk forever, but if they actually just see it, I think that’s going to speak
0:05:04 a thousand words, right?
0:05:04 Yeah.
0:05:05 Let’s jump into it.
0:05:06 So what are we going to do today?
0:05:09 So obviously, you know, the use cases that we do in the enterprise are, you know, obviously
0:05:12 very valuable, but a little less visually pleasing.
0:05:12 Right.
0:05:17 And so, you know, some common things that we do is like nasty migrations of like Java 7
0:05:23 to Java 21, database migrations, spring boot migrations, fun like SQL or COBOL.
0:05:27 Like all this stuff is very high value to the enterprise, but I think less maybe appealing
0:05:28 to the broad audience.
0:05:30 Yeah, I get like 50 views on YouTube or something.
0:05:31 Yeah, exactly.
0:05:33 So I was thinking, you know, for your audience, it might be fun.
0:05:37 People meme a lot about the number of engineers that work at DocuSign.
0:05:37 Oh, yeah.
0:05:40 It’s like 7,000 something crazy amount.
0:05:43 You know, every so often I’ll see a tweet that’s like after, you know, one company will announce
0:05:48 like a lot of layoffs, they’ll be like DocuSign announces hiring 100,000 more engineers
0:05:48 or something.
0:05:48 Yeah.
0:05:49 Yeah.
0:05:53 So I thought it could be fun to build a little toy version of DocuSign within Factory just
0:05:57 to get a sense of how Factory helps, you know, agentically automate some development
0:05:58 tasks.
0:05:58 Cool.
0:06:00 So we’re going to like build like a billion dollar company.
0:06:01 Yeah, let’s see what we can do.
0:06:03 Am I like a 50-50 partner in this or like how does this work?
0:06:06 Hey, if you help guide the droids, then you got to stay.
0:06:07 Okay, cool.
0:06:08 All right.
0:06:11 So here you can see we’ve landed on the Factory dashboard.
0:06:13 We have a chat interface.
0:06:15 So, you know, pretty standard, nothing too new here with LLMs.
0:06:17 We also have these four droids.
0:06:18 And so these are different.
0:06:19 These are our agents.
0:06:22 We call them droids because agent is kind of synonymous with poor quality.
0:06:23 Right.
0:06:25 It’s for, you know, different tasks that you might want to do.
0:06:28 So, you know, the knowledge droid is maybe you’re trying to create a design
0:06:31 dock for some large engineering tasks that your org will be doing.
0:06:33 Again, that’s a little bit more enterprise angle.
0:06:37 Code droid, this is really the kind of the one that everyone really wants to see, which
0:06:41 is, you know, whether it’s zero to one or end to end plus one, having the ability to go
0:06:44 into a code base and kind of build end to end features.
0:06:48 There’s the reliability drug, which can integrate with tools like Sentry or Datadog.
0:06:53 And if you have any outages, it can go in and create an RCA and even solve those issues.
0:06:56 We have the tutorial, which just guides you how to use Factory.
0:06:57 But cool.
0:07:00 I like the approach with the droids because it seems like you have these like specialized
0:07:05 droids that do a specific thing versus a lot of these, you know, AI agent companies.
0:07:09 I’m friends with Yohei Nakadima who did Baby AGI and actually like kind of spread that on
0:07:09 Twitter when it first came out.
0:07:11 And it was a great concept.
0:07:15 But a lot of these like that and Devin, they kind of promised they’re going to do everything
0:07:17 and then they kind of fail at almost everything.
0:07:21 I like the fact that you guys have like specialized ones that do specific things, have specific
0:07:22 deliverables.
0:07:23 Yeah, totally.
0:07:23 Yeah.
0:07:26 And I think, you know, the name of the game with agents is you want them to be as reliable
0:07:27 as possible.
0:07:32 And so if you could have them focused on certain core competencies in their workflows, it makes
0:07:34 that reliability a lot easier.
0:07:38 And so I just hit enter on, hey, let’s build out a toy version of DocuSign from scratch.
0:07:42 And so the code droid is now going in, creating this plan.
0:07:44 So, you know, here we have this general implementation plan.
0:07:49 Project setup, user authentication, document management, potential challenges.
0:07:53 Now, if I was going to be doing this a little more thoroughly and, you know, for production
0:07:56 grade right now, I might, you know, be a little more thoughtful in my responses.
0:07:58 But I defer to you.
0:08:04 Let’s just get a quick version up and running locally.
0:08:08 And then we can iterate from there.
0:08:13 So, you know, it’s asking me some very thorough questions like what tech stack, which core features
0:08:15 do you like this system to include authentication?
0:08:16 These are good questions.
0:08:17 I’m just saying, you know what?
0:08:18 You pick.
0:08:19 I don’t have time for this right now.
0:08:23 So, yeah, obviously in the enterprise setting, you’re going to be much more picky about this
0:08:25 when you’re not going, you know, purely from scratch.
0:08:29 But, you know, we can see it’s thinking process here as it’s going in and starting.
0:08:33 And so now it’s going to go and see here we have it running a command locally.
0:08:37 So the point here is that there are a lot of agents that run purely locally.
0:08:41 That’s like a tool like Cloud Code or any of the IDE agents.
0:08:42 They only run in your local environment.
0:08:47 And then there are some of the other agents like Codex or Devon, which only run remotely.
0:08:50 What’s incredible about Factory is we have the ability to do both.
0:08:52 And you can parallelize in both.
0:08:55 So as you’re delegating tasks, you can say some, you know what?
0:08:56 I’m confident in our tests.
0:08:58 I’m confident in our acceptance criterias.
0:09:00 I’m just going to go delegate that and send it to the cloud.
0:09:03 And then there are some where it’s like, I actually want to be pretty involved.
0:09:05 So I want this to work on it agentically, but locally.
0:09:09 Before you go on, I would love to like, so like, I believe I get what you said
0:09:10 in terms of being local and there’s a cloud.
0:09:13 And I think you said it, Greg Brockman was talking about this too, right?
0:09:16 That like that’s part of the vision in the future for OpenAI with Codex
0:09:17 is like that’s going to eventually do that.
0:09:20 And then you were telling me like basically Factory already does
0:09:22 what OpenAI is eventually doing.
0:09:22 Yeah, that’s right.
0:09:24 What’s the benefit of that kind of hybrid approach?
0:09:27 I mean, it allows these systems to have a silhouette
0:09:31 that’s much more similar to what we as humans have a silhouette.
0:09:32 So like, for example, when I’m working,
0:09:35 that looks like an agent working locally on my device.
0:09:37 Whereas when a colleague of mine is working,
0:09:41 it’s essentially equivalent to like some cloud environment,
0:09:42 you know, writing code.
0:09:44 And then how do I see the code that they write?
0:09:46 Well, they’ll submit a PR and then I’ll go and maybe look at their branch.
0:09:46 Right.
0:09:49 And so this way, it allows you to kind of spin up either
0:09:51 more copies of yourself or more copies of your colleagues.
0:09:52 Makes sense.
0:09:54 And so it’s kind of like, you know, a manager might think,
0:09:56 hey, here are these tasks that I’m going to do.
0:09:59 Here are these tasks that I’m going to kind of delegate to a colleague.
0:10:02 So you as kind of a pilot of this ship,
0:10:06 you kind of get to say, hey, which of these tasks will I monitor?
0:10:07 Which am I going to go send off?
0:10:09 And, you know, I’ll see the PRs later.
0:10:11 Yeah, so local is like giving you superpowers
0:10:13 or remote is like giving your team superpowers.
0:10:13 That’s right.
0:10:16 And with factory, you kind of get the best of both worlds.
0:10:16 It’s awesome.
0:10:17 Exactly right.
0:10:19 And so what happened here is, you know,
0:10:21 it was asking permission for certain commands.
0:10:23 I just turned on auto save and auto run.
0:10:26 So now it’s kind of just like full autonomous mode.
0:10:27 We’ll just like go run the commands.
0:10:30 So you can see it’s creating some repos, creating some folders.
0:10:33 And now it’s just up and running, you know,
0:10:35 and it’s going to create a few files
0:10:37 because you can’t make DocuSign just in one file.
0:10:40 So we can kind of have this running in the background here
0:10:41 and we can check in.
0:10:43 It should create all these files and then spin it up.
0:10:45 And so we’ll check it out.
0:10:47 Yeah, so this is basically like your YOLO mode or something?
0:10:48 Like it’s just like…
0:10:49 Basically, yes.
0:10:51 Except what’s nice is when you serve enterprise,
0:10:54 YOLO mode is not something that anyone ever wants.
0:10:55 Right, right, right.
0:10:56 We take this very seriously.
0:10:59 And so not only do you have this ability to just have like,
0:11:01 look, there are different levels of risk
0:11:03 for the auto accepting of CLI commands,
0:11:04 but also as an admin,
0:11:07 you’ll have the ability to whitelist or blacklist certain commands.
0:11:10 So you might not want to allow any pseudo commands
0:11:12 because it could do some pretty serious damage
0:11:14 or you might want to really restrict which folders
0:11:16 you’re even allowing the agent to get to.
0:11:18 How do you make that list of commands that can’t be accepted?
0:11:19 It’s in an admin setting.
0:11:20 Okay, cool.
0:11:21 Yeah, it’s just going to be running.
0:11:22 It might take a few minutes.
0:11:24 We can see it’s, you know, setting up its environment file.
0:11:26 But I think, you know, to your earlier question
0:11:28 about some of the things that I think OpenAI mentioned
0:11:29 in the codex launches,
0:11:32 it’s pretty clear that software development
0:11:35 is going to change dramatically over the next five years.
0:11:38 And I think an incongruence that currently exists
0:11:42 is that everyone says it’s going to change dramatically.
0:11:44 Yet the current paradigm,
0:11:49 pretty much what it looks like is putting AI onto existing workflows.
0:11:49 Right.
0:11:50 Right.
0:11:52 Like the existing workflow that developers have had for the last 15 years
0:11:54 is working in the IDE
0:11:56 and writing every single line of code there.
0:11:58 So we’ve applied AI into the existing workflows,
0:12:00 which is these AI IDEs.
0:12:02 But the reality is every big platform shift
0:12:05 has involved very significant behavior change.
0:12:07 You know, in the internet transition, what happened?
0:12:09 People went from getting most of their information from books
0:12:10 to like doing this,
0:12:11 and that’s how they get their information.
0:12:13 And mobile, what happened?
0:12:14 People went from like walking around with their heads up
0:12:16 to like walking around like that.
0:12:16 Yeah.
0:12:18 Very visceral behavior changes.
0:12:20 Yet AI, which is supposed to be the platform shift
0:12:22 that puts all these others to shame,
0:12:23 what are the most used products?
0:12:27 Well, it’s like ChatGPT, Perplexity, Copilot, Cursor.
0:12:28 Well, ChatGPT and Perplexity,
0:12:30 that’s just Google with better results.
0:12:31 Right.
0:12:31 Right.
0:12:32 It’s the same behavior.
0:12:33 Chat with Google, basically.
0:12:34 Exactly.
0:12:36 And then similarly with like Copilot and Cursor,
0:12:37 it’s essentially the same behavior,
0:12:39 which is like the IDE behavior.
0:12:40 Now you’re maybe pressing
0:12:42 some slightly different keys more often,
0:12:44 but it’s not like a viscerally changed behavior.
0:12:45 Right.
0:12:46 And that’s because we haven’t hit
0:12:48 that full transformation yet.
0:12:50 And what we’re really focused on with our droids
0:12:53 and with the ability to have them local and remote
0:12:55 is this is the new behavior that’s going to emerge
0:12:58 where you’re not writing every line of code.
0:13:00 the center of gravity of software development
0:13:05 will change from coding to instead understanding and planning
0:13:07 and then testing to make sure that these agents,
0:13:09 when they go and submit their PRs,
0:13:11 they did satisfy the constraints that you had in mind.
0:13:13 Yeah, that makes a lot of sense to me.
0:13:14 I like the guy, Primogen.
0:13:15 Do you know him on YouTube?
0:13:16 Yeah, of course.
0:13:17 Yeah, yeah.
0:13:17 He’s awesome.
0:13:19 And he’s slightly warming up to AI now,
0:13:21 but at first he was really hating on AI.
0:13:22 And it’s just because he just, you know,
0:13:25 he’s passionate about the art of just coding itself.
0:13:26 I’m lightly technical,
0:13:27 but I’m more of a business person, you know, investor.
0:13:29 And I think about it as like, you know,
0:13:30 when engineering, you know,
0:13:32 the point is to create things, right?
0:13:33 To solve problems.
0:13:34 Right.
0:13:35 And that’s what engineering really is about.
0:13:37 But even if we love it, you know,
0:13:38 that doesn’t mean that’s the best way
0:13:40 to do things in the future, right?
0:13:41 That’s why when I saw Factory’s website,
0:13:43 I was like, this feels more like the future.
0:13:44 100%.
0:13:46 And also to your point, it’s about building things.
0:13:46 Yeah.
0:13:47 So what does it mean to build something?
0:13:48 Well, you have an idea.
0:13:48 Yeah.
0:13:50 And that idea is defined by some constraints.
0:13:52 Let’s say your idea was Spotify.
0:13:53 It’s like, okay, well,
0:13:55 you have this social music sharing app.
0:13:57 Obviously there’s a business side of like
0:13:59 having all the agreements with the record labels,
0:14:01 but from the product itself,
0:14:01 it’s like, you know,
0:14:04 some low latency, high fidelity music sharing
0:14:06 that has social features, playlists, all that stuff.
0:14:08 Well, those are certain constraints
0:14:09 that you have in your head.
0:14:09 Right.
0:14:11 And what you need to do to actually build that
0:14:13 is you need to turn those constraints
0:14:16 into machine readable language,
0:14:17 which is, that’s why we have these,
0:14:18 you know, programming languages.
0:14:21 But what’s getting unlocked now
0:14:23 is that translation from you to the computer.
0:14:25 It used to be, you know,
0:14:27 you needed however many years of an education
0:14:28 and years of experience to actually learn
0:14:29 how to do that translation.
0:14:31 You get to hire people and like hope
0:14:32 that they actually did the work
0:14:33 and actually did what they said they were doing.
0:14:34 Maybe raise money.
0:14:34 Yeah.
0:14:35 Also, it’s like, that’s difficult.
0:14:37 Not a lot of people have access
0:14:38 to be able to hire that many people
0:14:39 or to raise the capital to do that.
0:14:40 Right.
0:14:40 Whereas now,
0:14:43 if you’re able to translate those constraints
0:14:44 that you have in your head,
0:14:47 you can kind of speak to a tool like Factory
0:14:49 and it will translate those constraints
0:14:50 into the software itself.
0:14:52 And so it lowers so many of those barriers
0:14:54 and kind of, again, refocuses like
0:14:55 what has made the best engineers
0:14:57 and the best product thinkers.
0:14:59 It’s not that they know every little detail
0:15:00 about every little language,
0:15:01 but they’re the best at thinking
0:15:02 about those constraints
0:15:05 and understanding what does my customer want
0:15:06 and how do I translate that
0:15:07 into these constraints?
0:15:07 Yeah, totally.
0:15:08 Thinking more about the customer
0:15:09 and spending more time on that
0:15:10 and the experience
0:15:11 and what problems you’re solving
0:15:13 versus dealing with bugs.
0:15:13 Yeah.
0:15:14 You know, I was in Silicon Valley
0:15:15 for 13 years.
0:15:16 I coded on and off.
0:15:17 I never like was super hardcore
0:15:18 into coding,
0:15:19 but a lot of my friends were
0:15:20 and I was just,
0:15:21 you know, I just felt like,
0:15:21 God, it’s, you know,
0:15:23 I just want to like solve problems
0:15:24 for people and create cool stuff,
0:15:24 right?
0:15:25 And tell people about it.
0:15:27 And like every time I started coding,
0:15:28 it was just like bugs
0:15:29 would just like drive me crazy.
0:15:30 I’m like, why am I like spending my life
0:15:31 dealing with these stupid bugs?
0:15:32 And then what other people,
0:15:34 you know, talk about like that.
0:15:34 It’s like the great thing
0:15:35 to be working on bugs.
0:15:36 Like, Jesus, it’s not.
0:15:37 You got one life.
0:15:38 Why would you spend all your life
0:15:39 solving bugs?
0:15:40 You want to build things?
0:15:41 Yeah, you’re right.
0:15:42 Right.
0:15:43 It’s like no one enjoys doing it.
0:15:44 And that’s the equivalent of that
0:15:45 in like the enterprise is like,
0:15:47 there’s so many menial,
0:15:48 tedious tasks
0:15:50 that enterprise engineers have to do.
0:15:50 Right.
0:15:51 It’s not why they got into engineering
0:15:52 in the first place.
0:15:52 Right.
0:15:53 And even something that you mentioned
0:15:55 that I think is a really interesting point to me
0:15:57 is you want to build something cool
0:15:57 or make something cool.
0:15:58 But the thing is,
0:16:00 you can’t just go to an LLM and say,
0:16:01 hey, make me something cool.
0:16:04 Because when you say cool in your head,
0:16:06 there are certain things that that means
0:16:07 that the LLMs don’t know.
0:16:10 And so this new era of software developers
0:16:12 who are the like best going to be,
0:16:14 the best people will be the ones that
0:16:15 when they think cool,
0:16:16 they know, like,
0:16:18 how do you elicit that from the model?
0:16:18 Right.
0:16:19 Because if I just said,
0:16:20 make me something cool,
0:16:22 probably I’m not going to be happy with it.
0:16:22 Right.
0:16:23 Because in my head,
0:16:24 when I thought cool,
0:16:25 I really meant something else.
0:16:26 Right.
0:16:26 Right.
0:16:26 Yeah.
0:16:27 Yeah.
0:16:28 And this is perfect for people like me
0:16:30 because I find like when I talk,
0:16:31 I’m a lot less articulate than I write.
0:16:33 I love to like sit around for an hour or two
0:16:34 and think something through
0:16:35 and write it out
0:16:36 and maybe very precise.
0:16:37 Yeah.
0:16:38 I mean, I used to be a physicist.
0:16:40 I spent a lot of my time with mathematicians
0:16:41 and there’s so many people
0:16:43 who felt more comfortable
0:16:45 explaining ideas with math
0:16:46 than with words
0:16:47 because sometimes like words
0:16:48 are very difficult
0:16:50 and similarly like engineers,
0:16:51 like sometimes it’s very difficult
0:16:53 to describe what you mean in words
0:16:54 as opposed to in like,
0:16:55 you know, programming language.
0:16:57 But understanding how to speak that
0:16:58 to the models
0:16:59 without writing out all the code yourself,
0:17:01 I think that’s going to be
0:17:02 a very powerful skill.
0:17:03 It’s kind of awesome
0:17:04 that we’re having this conversation right now
0:17:06 while like a billion dollar company
0:17:07 is being built in the background.
0:17:08 Yeah, exactly.
0:17:09 Yeah.
0:17:10 Hopefully it works.
0:17:12 I’m always nervous for like founders
0:17:13 when they do a live demo with AI
0:17:13 because, you know,
0:17:15 you never know what’s going to happen.
0:17:16 It looks like it’s doing
0:17:17 roughly the right thing.
0:17:18 It’s making some layouts,
0:17:18 some loading spinner,
0:17:20 login, register, dashboard,
0:17:21 uploading document,
0:17:22 creating signature.
0:17:22 That’s good.
0:17:23 You know,
0:17:24 got to have signatures
0:17:24 and DocuSign.
0:17:25 So yeah,
0:17:26 it seems like it’s on the right track.
0:17:30 Hey, we’ll be right back to the show.
0:17:30 But first,
0:17:31 I want to tell you about another podcast
0:17:33 I know you’re going to love.
0:17:35 It’s called Marketing Against the Grain.
0:17:36 It’s hosted by Kip Bodner
0:17:38 and Kieran Flanagan.
0:17:39 And it’s brought to you
0:17:40 by the HubSpot Podcast Network,
0:17:42 the audio destination
0:17:43 for business professionals.
0:17:44 If you want to know
0:17:45 what’s happening now in marketing,
0:17:47 especially how to use AI marketing,
0:17:49 this is the podcast for you.
0:17:50 Kip and Kieran
0:17:52 share their marketing expertise
0:17:53 unfiltered in the details,
0:17:54 the truth,
0:17:55 and like nobody else
0:17:56 will tell it to you.
0:17:58 They recently had a great episode
0:18:00 called Using ChatTBT 03
0:18:03 to Plan Our 2025 Marketing Campaign.
0:18:05 It was full of like actual insights
0:18:06 as well as just
0:18:08 things I had not thought of
0:18:11 about how to apply AI to marketing.
0:18:13 I highly suggest you check it out.
0:18:15 Listen to Marketing Against the Grain
0:18:16 wherever you get your podcasts.
0:18:20 I mean, it seems like the fact
0:18:21 that you were like tapping
0:18:23 into like the existing systems
0:18:23 inside of companies
0:18:24 where they already,
0:18:26 you know, use ticketing systems
0:18:26 and everything else,
0:18:27 like it seems like
0:18:29 a more natural fit for enterprises.
0:18:30 And I think it makes a lot of sense.
0:18:31 You know, we talk about this
0:18:32 in the podcast a lot.
0:18:33 Like I had Greg Eisenberg on
0:18:33 and we talked about like
0:18:34 what’s the future of business
0:18:35 and things like that.
0:18:37 And it feels like a lot of companies,
0:18:38 they’re starting to wake up to this,
0:18:40 but in the AI age,
0:18:41 it’s going to be more and more important
0:18:42 to keep reinventing yourself,
0:18:43 right, as a company.
0:18:44 And if you don’t,
0:18:45 a lot of companies
0:18:46 are going to be out of your business.
0:18:46 I mean, I think we’re going to see
0:18:49 lots of like multi-trillion dollar companies
0:18:50 that people are going to be shocked.
0:18:51 They thought trillion was a big deal.
0:18:52 It’s going to seem like
0:18:53 a small number in the future.
0:18:54 And we’re going to see
0:18:55 a lot of companies
0:18:56 that you thought were going to be around
0:18:57 that are no longer going to be around.
0:18:57 Yeah.
0:18:57 And it’s because
0:18:58 they didn’t reinvent themselves.
0:18:59 And, you know,
0:19:00 if I was a big company,
0:19:01 I would be looking at something like this
0:19:02 and thinking like,
0:19:03 this is something I can actually
0:19:04 get my teams to use.
0:19:05 It fits into our existing systems
0:19:07 and it allows us to
0:19:08 make our teams happier
0:19:09 to actually do work they want to do
0:19:10 and to try new things
0:19:11 because they need to be
0:19:12 trying new things right now.
0:19:13 Yeah.
0:19:14 You know, it is a little bit scary
0:19:15 to think about like,
0:19:16 oh, there are a lot of companies
0:19:16 that, you know,
0:19:17 if they don’t reinvent themselves,
0:19:18 there are certain things
0:19:19 that become obsolete.
0:19:20 On a micro scale,
0:19:21 it’s a little alarming
0:19:22 or a little concerning.
0:19:23 But I think on a macro scale,
0:19:24 it’s happening
0:19:25 because things are becoming
0:19:26 more efficient.
0:19:26 Right.
0:19:27 which I think is actually
0:19:28 a great story, right?
0:19:29 Like, I mean,
0:19:31 something that we talk about a lot,
0:19:32 even when we build factories,
0:19:33 the Henry Ford quote,
0:19:33 which is,
0:19:35 if you ask people what they want,
0:19:36 they would say faster horses.
0:19:38 And sometimes you need to kind of
0:19:39 look past what people are asking for
0:19:40 and build that automobile.
0:19:40 Right.
0:19:41 And something that,
0:19:42 you know, is a big change
0:19:44 is like in a world with horses,
0:19:45 the structure of the economy
0:19:46 is very different
0:19:47 than a world with cars.
0:19:47 Right.
0:19:48 But net net,
0:19:50 the world is much more efficient.
0:19:51 And so short term,
0:19:52 there are things that change,
0:19:53 but long term,
0:19:53 you know,
0:19:55 people who want to visit their families,
0:19:56 they’re able to do so faster.
0:19:56 Right.
0:19:57 If you have a medical emergency,
0:19:59 you can get to a hospital faster.
0:20:00 And so it’s like,
0:20:01 sometimes when you have these
0:20:03 step function changes in efficiency,
0:20:05 like the way the world has been built
0:20:07 will change a little.
0:20:07 Right.
0:20:07 But net,
0:20:09 it’s kind of a narrative of like
0:20:10 things becoming more and more efficient.
0:20:12 And I think that’s always a good thing.
0:20:13 And the creative people,
0:20:13 the smart people,
0:20:14 the great engineers,
0:20:15 the great product thinkers
0:20:17 will get to work on higher leverage,
0:20:18 more efficient problems,
0:20:20 which I think is a net good for the world.
0:20:20 Yeah.
0:20:22 And I agree with the smarter people
0:20:22 getting in from this a lot,
0:20:24 but I also think people who were
0:20:25 just like average
0:20:26 will also do quite well
0:20:28 because as long as you’re very persistent
0:20:29 and willing to grind and hustle,
0:20:30 like,
0:20:31 I think this is going to be amazing
0:20:32 for those people
0:20:32 because like maybe before
0:20:34 they weren’t the best coder
0:20:36 and they’d go talk to some engineer
0:20:36 and be like,
0:20:36 hey,
0:20:37 here’s my idea.
0:20:39 The person would be in the back of their head
0:20:39 and go,
0:20:39 okay,
0:20:40 cool idiot,
0:20:40 you know,
0:20:41 and like actually judging them
0:20:43 based on what they judged their IQ to be
0:20:43 or whatever.
0:20:44 Yeah.
0:20:45 but now with tools like this,
0:20:47 they can just go talk to the AI
0:20:47 and build the thing
0:20:49 and spread it out there into the world,
0:20:51 which wouldn’t have been possible before.
0:20:52 So that’s super exciting to me.
0:20:52 Totally.
0:20:53 And I think that’s also,
0:20:54 it kind of hits on something
0:20:55 that I found very compelling,
0:20:57 which is a lot of people are saying,
0:20:57 oh,
0:20:58 you know,
0:20:59 all these models are kind of
0:21:01 commodifying IQ or intelligence.
0:21:01 Yeah.
0:21:02 So that now like,
0:21:04 it doesn’t matter if you’re not that smart
0:21:06 because whatever the model at your fingertips,
0:21:07 you can have whatever intelligence you need.
0:21:09 And so then the question is,
0:21:09 okay,
0:21:09 well,
0:21:11 a prior theory is that your success
0:21:13 or your ability to do well
0:21:14 is kind of defined by,
0:21:14 you know,
0:21:16 IQ and how hard you can work
0:21:16 and that sort of thing.
0:21:18 And in this world where like,
0:21:19 this work is via API
0:21:21 or this intelligence is via API,
0:21:22 what is this new metric
0:21:24 that will determine like success
0:21:25 or not success?
0:21:27 Something that I think is kind of interesting
0:21:29 is the idea of agency
0:21:30 being the determining factor.
0:21:32 So even if you’re not the highest IQ,
0:21:34 if you have the will to go
0:21:35 and build things
0:21:37 as opposed to kind of being passive and lazy,
0:21:39 that’s going to be the determining factor.
0:21:40 And if you have that agency,
0:21:42 you can go and use all these models
0:21:44 that have that intelligence on demand for you
0:21:45 that might be experts
0:21:46 in all these niche fields
0:21:47 that you don’t have time
0:21:48 to become an expert in.
0:21:49 I find that kind of interesting.
0:21:50 Yeah,
0:21:50 I think that’s right.
0:21:52 I still think IQ is going to be a big advantage,
0:21:52 honestly.
0:21:54 I think this is going to amplify people
0:21:55 who are high IQ.
0:21:57 Maybe they’ll get ahead even further.
0:21:57 I mean,
0:21:58 a question is how closely
0:22:00 is agency tied to IQ,
0:22:01 which I think is an interesting question.
0:22:02 Right, yeah.
0:22:03 Yeah.
0:22:04 That’s a whole other conversation.
0:22:04 Yeah.
0:22:05 Cool.
0:22:06 So it’s still working.
0:22:07 It’s creating a lot of files, yeah.
0:22:08 Yeah, one thing I was thinking about
0:22:09 is one of our first guests
0:22:11 was Ervin Srinivas of Perplexity.
0:22:12 Oh, nice.
0:22:13 And when he first came on,
0:22:14 I was worried that, you know,
0:22:15 OpenAI was just going to eat them
0:22:16 like very quickly and destroy them.
0:22:16 Yeah.
0:22:17 Hasn’t happened.
0:22:18 They’ve done incredibly well.
0:22:20 And I feel like one of the benefits
0:22:21 they’ve had is the fact that,
0:22:22 you know,
0:22:23 every month there’s a different model
0:22:24 that becomes the best model.
0:22:27 And they’re able to tap into that
0:22:28 and take advantage of that
0:22:28 versus,
0:22:29 okay,
0:22:29 it’s just ChatsBT.
0:22:30 It’s just, you know,
0:22:31 Claude or whatever.
0:22:32 I was sitting there thinking like,
0:22:33 well,
0:22:34 Google’s got jewels now.
0:22:35 OpenAI has codex.
0:22:36 It feels like the fact
0:22:37 that you guys can tap into
0:22:38 whatever model is the best currently
0:22:40 is a huge advantage.
0:22:41 Totally.
0:22:42 All right.
0:22:44 So we look like we are up
0:22:45 and running here.
0:22:46 So I’m just going to ask
0:22:47 to set this guy up.
0:22:48 We’re still connected
0:22:49 to my local machine.
0:22:50 So looking good.
0:22:51 And we should be able to
0:22:53 check out our little DocuSign.
0:22:54 Let’s see how it looks.
0:22:55 Cool.
0:22:56 Fingers crossed.
0:22:58 All right.
0:23:00 So let’s check it out locally.
0:23:03 And we are good.
0:23:04 Oh my God.
0:23:06 This is our little DocuSign toy.
0:23:07 You got a landing page.
0:23:07 I mean,
0:23:08 does it do anything?
0:23:08 I mean,
0:23:09 that’s,
0:23:09 yeah,
0:23:10 we got a landing page.
0:23:10 The colors,
0:23:11 you know,
0:23:11 could do some work.
0:23:12 Very blue.
0:23:14 Not sure how I feel about this logo,
0:23:14 but you know what?
0:23:16 We asked for DocuSign.
0:23:16 Let’s see.
0:23:17 Sure,
0:23:18 DocuSign would be fine with that name.
0:23:19 DocuSign toy.
0:23:19 Why not?
0:23:20 DocuSign toy,
0:23:20 yeah.
0:23:20 All right.
0:23:22 So let’s try a little login.
0:23:22 So again,
0:23:24 we see we’re missing some of these logos,
0:23:25 but we’ll get our designer in here
0:23:26 to make it look cooler.
0:23:28 But I think the demo account
0:23:31 is user at example.com.
0:23:32 Password’s password.
0:23:35 And it looks like we’re in.
0:23:37 It seems pretty sweet.
0:23:38 Jesus.
0:23:38 Yeah.
0:23:39 So,
0:23:39 you know,
0:23:40 we can make some templates.
0:23:41 We can take a tour.
0:23:42 Let’s upload a template.
0:23:43 Okay.
0:23:45 Let’s throw in a PDF.
0:23:48 Let’s throw in a factory’s one pager.
0:23:48 Yeah.
0:23:51 and I’ll send it to myself.
0:23:53 Upload the Doc.
0:23:53 Okay.
0:23:54 Looks like a Doc.
0:23:55 Let’s add a signature.
0:23:56 There we go.
0:23:59 Add an initial field.
0:23:59 There we go.
0:24:01 Let’s send it for signature.
0:24:04 And we can see I now have one awaiting signature over here.
0:24:06 So,
0:24:06 yeah,
0:24:08 I don’t know about this layout necessarily,
0:24:08 but, you know,
0:24:09 we can adjust it.
0:24:09 Okay.
0:24:11 Send me the Docs to own half of this,
0:24:11 okay?
0:24:12 We’ll launch it.
0:24:13 All right.
0:24:14 So,
0:24:14 here we see,
0:24:15 so I have one to sign.
0:24:15 So,
0:24:16 let’s go and sign it.
0:24:18 Click to sign.
0:24:19 Can add my signature.
0:24:21 Click to initial.
0:24:22 Add the initials.
0:24:23 Seems pretty legit.
0:24:24 Complete signing.
0:24:25 There we go.
0:24:26 And now we have one completed.
0:24:27 Not too bad.
0:24:28 That is amazing.
0:24:29 Oh,
0:24:30 and we have a little activity log here as well.
0:24:31 That’s pretty sweet.
0:24:33 Let’s go back to the dashboard.
0:24:34 Okay.
0:24:35 That is incredible.
0:24:36 Okay.
0:24:36 So,
0:24:39 Factory is amazing at coding and creating a SaaS app.
0:24:41 It overuses blue.
0:24:42 There we go.
0:24:42 Hey,
0:24:43 you know,
0:24:43 we can make it red.
0:24:44 And that’s the thing is,
0:24:44 you know,
0:24:46 we can go back into Factory and adjust from there.
0:24:47 I think,
0:24:47 honestly,
0:24:51 the first thing that would just make this look a lot nicer is all these like placeholder icons.
0:24:52 Right.
0:24:53 Like having those,
0:24:54 that would make this look pretty sweet.
0:24:54 Yeah.
0:24:56 Here we even have templates already.
0:24:56 Yeah.
0:24:59 I think a big thing would be maybe adjusting some of the ordering here.
0:24:59 Like,
0:25:03 I think this is a little too much scroll when there’s nothing populated in there.
0:25:04 But honestly,
0:25:07 the only thing missing for this is like hooking it up to a database,
0:25:08 like to a proper backend.
0:25:08 Right.
0:25:12 And then all of the compliance stuff that I’m sure DocuSign actually has to deal with.
0:25:14 But you probably could talk to a factory and get a lot of that done.
0:25:16 Maybe not all of it at DocuSign level,
0:25:18 but like close enough.
0:25:18 Right.
0:25:19 For one shot in 15 minutes,
0:25:20 this is a very,
0:25:21 very solid start.
0:25:22 Yeah.
0:25:22 Yeah.
0:25:24 So like if you’re an entrepreneur watching this,
0:25:24 I mean,
0:25:25 in theory,
0:25:25 you know,
0:25:26 DocuSign is like,
0:25:28 I think they’re like an $8.6 billion company.
0:25:29 Right.
0:25:30 Yeah.
0:25:30 So in theory,
0:25:33 you could create something at that level as a,
0:25:34 even a one person team.
0:25:35 Which is just mind blowing.
0:25:36 Pretty exciting to see.
0:25:43 And I think it’s really also cool because it just means that barrier to creating that next $8 billion company is that much lower.
0:25:44 Right.
0:25:45 because now instead of the,
0:25:46 let’s say 500 engineers,
0:25:47 you might’ve needed to make that,
0:25:48 you know,
0:25:48 you know,
0:25:49 kidding.
0:25:49 But you know,
0:25:52 it just allows you to build these things out much faster.
0:25:53 And even within the enterprise,
0:25:54 in large orgs,
0:25:58 there are a lot of times where there’ll be teams of like 20 building out internal tools.
0:26:02 And it’ll take them just a huge amount of time when really the internal tool is a means to an end.
0:26:02 Right.
0:26:03 It allows you to get to that end faster.
0:26:04 Very cool.
0:26:10 You’ve been serving the enterprises for the last year or two and now it’s generally available.
0:26:11 Is that what’s happening right now?
0:26:11 That’s right.
0:26:11 Yeah.
0:26:12 We’re fully GA.
0:26:14 We have a team’s plan now.
0:26:16 That’s just to start $40 a month.
0:26:18 You can invite other teammates for an additional,
0:26:19 at least for now,
0:26:21 $10 for every additional user.
0:26:21 That’s all?
0:26:23 That’s crazy.
0:26:23 That’s all.
0:26:24 That is all.
0:26:24 Okay.
0:26:25 Yeah.
0:26:25 I mean,
0:26:28 I think the thing that we’re seeing is just a lot of small teams are really liking this.
0:26:29 And,
0:26:29 you know,
0:26:31 our focus still remains on the enterprise,
0:26:33 but if there’s something that delivers value to,
0:26:34 you know,
0:26:36 a demographic that’s slightly different than what we initially targeted,
0:26:39 initially there was a reason for the sake of focus to not kind of open that up.
0:26:41 But now we have the scalability.
0:26:43 We just want to put this in more people’s hands.
0:26:44 It’s still early.
0:26:45 I’m sure there are going to be things that,
0:26:46 you know,
0:26:47 people have a lot of feedback.
0:26:49 There might be things that they want to adjust and we’re very eager to hear.
0:26:50 But yeah,
0:26:52 I’m excited to put it in more people’s hands.
0:26:52 That’s awesome.
0:26:53 I see where it’s at today.
0:26:54 I think it has so much potential.
0:26:55 I mean,
0:26:56 where do you think factory is going to go in the future?
0:26:56 Like,
0:26:58 let’s say this is a super successful,
0:26:58 you know,
0:27:00 GA launch and everyone loves it.
0:27:02 And when people start talking about Devin,
0:27:05 they now say factory and you become part of like one of the category leaders.
0:27:07 Like where does factory go in the next like three to five years?
0:27:08 Yeah.
0:27:08 I mean,
0:27:14 I think the big thing in the near horizon is being this kind of unified platform for software development.
0:27:14 Right now,
0:27:17 a developer kind of lives a very fragmented life between GitHub,
0:27:18 between their IDE,
0:27:19 between Slack,
0:27:21 between Google drive,
0:27:22 between notion linear,
0:27:23 all these different tools.
0:27:29 And in a similar way to how a startup like Rippling kind of unified HR and IT into one place.
0:27:38 That’s what we want to do with software development because there’s just so much time spent crawling between all these different platforms and kind of pulling in all that information.
0:27:48 And that slows down that journey from idea to feature and factory and our droids are going to have access to all of these tools and they’ll meet you wherever you need.
0:27:54 So just like anything that’s like kind of most convenient is you need to do the least in the way of getting there.
0:27:55 And I think for us,
0:27:57 we want to meet developers where they are in these existing tools,
0:28:04 but then also provide them this nice new comfy home within factory where they can even start their projects as well.
0:28:09 So whether you have a very long Slack thread that’s an important conversation about product feature that you want to build out,
0:28:10 you can then tag factory.
0:28:16 It’s going to go and start creating a first pass, either like design doc or even a PR based on that.
0:28:21 Or if you have a backlog to tickets in your linear or JIRA, tag a droid and it’ll go and submit a PR to solve it.
0:28:22 That’s amazing.
0:28:25 So like, like even executives could be talking in Slack and like, you know,
0:28:28 they have an idea versus like going and bothering the engineering team.
0:28:32 They can just like have a first pass at it and see if it’s actually close to what they were imagining.
0:28:34 And then maybe, then maybe handing off the engineering team to take it to the next level.
0:28:35 A hundred percent.
0:28:39 I mean, something we’ve seen like that kind of naturally emerged that we weren’t expecting is
0:28:43 PMs at some of these enterprises we deployed to kind of got their hands on factory.
0:28:49 We didn’t initially plan on deploying to them and it kind of raised the bar for what is like a demo or a proof of concept internally.
0:28:57 Or also, you know, there’s so many times where PMs would need to ping front end engineers or full stack engineers to change copy or to add a page or this or that,
0:29:00 which like just slows down the engineering org so much.
0:29:01 They hate it.
0:29:04 It’s like, let me go in there and just change some stuff that you could just do if you knew what the hell you were doing.
0:29:05 Exactly.
0:29:05 Exactly.
0:29:07 And so now they don’t need to do that at all.
0:29:10 And they’re bragging like, hey, look, I just shipped some production ready PRs.
0:29:14 And if you’re an engineering leader and you’re worried like, oh no, my PMs are going to start submitting a lot of PRs.
0:29:17 What’s great about factories, we also adhere to your best practices.
0:29:26 So if you have pretty thorough docs in your org about like, here’s how we write tests, here’s how we ship features, here’s like our contributing guides, the droids will adhere to that.
0:29:32 And so if someone’s trying to ship a PR, it’ll actually go in and make the changes to make it adhere to whatever standards that you have.
0:29:33 That’s awesome.
0:29:39 And so it kind of keeps in check and makes sure you’re not just like introducing a lot of kind of vibe coded PRs in there.
0:29:39 Right.
0:29:42 But, you know, adhere to the enterprise standards, yeah.
0:29:45 Yeah, somewhat contributed to the vibe coding trend with Rally Brown.
0:29:46 I think it’s cool.
0:29:48 I mean, I think for like simple little apps, it’s cool.
0:29:51 But I think for anything complex, it starts to break down.
0:29:57 And I think it’s brilliant how you guys have started with enterprises and now kind of work to where regular people can use it as well.
0:30:04 I think also companies like OpenAI, I don’t see them anytime soon, like building something for the enterprise where they build everything out that the enterprise would need.
0:30:04 Right.
0:30:08 They’re going to start with more consumers because I mean, ChatsBT is a consumer app as of now.
0:30:08 Yeah.
0:30:10 So I love the strategy.
0:30:12 You keep saying droids, like why?
0:30:17 I mean, I know you kind of touched on it earlier, but like, are you not concerned about George Lucas or anything like that?
0:30:18 Yeah, no.
0:30:22 So we actually initially were incorporated as the San Francisco droid company.
0:30:23 Yeah.
0:30:23 Okay.
0:30:32 We were advised by our lawyers that Lucasfilm is very litigious and we decided to rename to the San Francisco AI Factory.
0:30:34 But we really loved the name.
0:30:36 Honestly, our customers really love the name too.
0:30:40 I can’t tell you how many times people are like, I speak with the droids or something like that.
0:30:47 And so now I think it’s more just, it’s going to be a sign of success when we get our first cease and desist from Lucasfilm.
0:30:48 Crazy side story.
0:30:53 So I don’t know George, but actually the reason I have lore.com is I was partnered with Barry Osborne, the producer of Lord of the Rings and the Matrix.
0:30:55 We were trying to make movies do it together.
0:30:55 Oh, no way.
0:30:57 And yeah, it was crazy for me.
0:31:03 Like I was involved in crypto pretty early on and like I sold my startup, not for like a huge amount, but you know, I was kind of in between projects.
0:31:07 My buddy introduced me to Barry and I was like, wait, you want to work with me on this stuff?
0:31:08 And like, let’s try to do it together.
0:31:09 We like really hit it off.
0:31:10 Damn.
0:31:15 You know, almost became like a somewhat like almost like a father figure to me, even though we’re like business partners, you know.
0:31:19 And he started getting me involved in meeting all these amazing people in Hollywood.
0:31:22 And I was out in Japan and he messages me.
0:31:26 He’s like, hey, do you want to go to a Skywalker ranch and possibly meet George?
0:31:26 That’s crazy.
0:31:29 I was like, yes, yes, I want to go.
0:31:33 And I literally went from Japan back to San Francisco.
0:31:35 You know, I don’t know anything about Hollywood.
0:31:35 I grew up in Alabama.
0:31:37 I bought this fancy jacket and everything.
0:31:39 I thought it was gonna be cool going to hang out out there.
0:31:42 I fly back from Japan to San Francisco.
0:31:46 And then that was when the wildfires happened and the entire thing got canceled.
0:31:46 Oh, no.
0:31:47 The entire thing got canceled.
0:31:50 And so it just never happened again.
0:31:51 And then COVID happened right after that.
0:31:54 And so I never got to meet those people, which was really sweet.
0:31:55 Oh, man.
0:31:55 That would have been awesome.
0:31:59 So something I think, you know, I just mentioned my son, because actually he went back to San
0:32:00 Francisco with me.
0:32:04 One thing I think about a lot is I think about, you know, what should I be teaching my son?
0:32:07 In the AI age, should he be learning to code?
0:32:09 It’s hard to know what he’s going to do in the future, right?
0:32:09 Yeah.
0:32:11 You’re too young to have a kid right now, or probably are.
0:32:13 If you did have a kid, would you be teaching them to code?
0:32:14 Yeah.
0:32:16 I mean, I think 100% unequivocally.
0:32:23 Just like I think, you know, even though I don’t need to do multiplication that often these
0:32:27 days, I think understanding the things that underlie all of the technology around us will
0:32:28 always be important.
0:32:28 Right.
0:32:33 I think similarly, like understanding like machine code doesn’t really matter or like assembly
0:32:34 doesn’t really matter.
0:32:39 But to have that full kind of systems understanding of the different layers of abstraction will always
0:32:39 be important.
0:32:39 Right.
0:32:44 Whether you’re a software engineer, a product builder, whether you’re a theoretical physicist,
0:32:48 you’re like, it’s still important to understand kind of the bare bones of what underlies
0:32:50 whatever it is that you’re working on.
0:32:55 And I think what we’re coming to terms with is there’s kind of going to be a mountain of
0:33:00 material that we no longer need to know, but it’s still like you will be at a huge advantage
0:33:01 if you are familiar with that.
0:33:02 Yeah.
0:33:03 So I think coding is incredibly important.
0:33:07 I think mostly for the way it teaches you how to reason and how to think.
0:33:07 Right.
0:33:08 That’s what I was going to say.
0:33:10 In that systems way, thinking about constraints.
0:33:14 Again, I think that methodical way of thinking and reasoning through problems, that’s always going
0:33:15 to be valuable.
0:33:20 I was a physicist before, and there’s a funny thing where there are a lot of physicists in
0:33:22 a lot of the foundation model labs.
0:33:25 And is it because there’s a lot of black holes involved in the LRMs?
0:33:25 No.
0:33:26 Not as far as we know.
0:33:27 Yeah.
0:33:27 As far as we know.
0:33:28 I don’t know.
0:33:28 Yeah.
0:33:29 Maybe this is maybe this laid poorly.
0:33:35 But the reality is working on problems that have very difficult reasoning and require the
0:33:39 synthesis of a lot of different information and reasoning about it in a pretty like non-trivial
0:33:44 quantitative way and kind of a systems way, that is just a valuable skill no matter what
0:33:45 and whatever domain you end up applying it to.
0:33:46 Right.
0:33:48 And so I think this applies similarly with these tools.
0:33:52 Now, I think this next generation should not necessarily be brought up the same way we
0:33:55 were because they should also be native in how to use these tools.
0:34:01 Just like how in eras before, people would spend a lot of time with an abacus doing calculations.
0:34:04 It’s important to know, OK, I can do the calculation myself.
0:34:07 But once you’re past that, it’s like, OK, now use a calculator from now on.
0:34:12 Just remember, you still know how to do the math because now you can use it and have much
0:34:14 higher leverage to kind of build things.
0:34:15 I agree.
0:34:18 You know, I’ve seen a lot of people, at least on Twitter from Silicon Valley, like commenting
0:34:22 on this and they seem to have like really, I don’t know, they treat it too binary in my
0:34:22 opinion.
0:34:27 Like I’ve seen either like, yeah, my kid’s like seven and they’re like going through like a
0:34:28 coding boot camp.
0:34:31 Or it’s like, I’m never teaching my kid to code because it doesn’t matter anymore.
0:34:34 Neither one of these makes sense to me right now.
0:34:37 Like with my son, I’ve been like kind of easing him into things like I showed him the command
0:34:37 line.
0:34:38 He’s like, oh, that’s cool.
0:34:39 That’s how that works.
0:34:41 I’m like, yeah, this is actually what’s going on behind the scenes.
0:34:42 Yeah.
0:34:44 I taught him like four or five commands.
0:34:46 I was like, if he wants to play with it sometime, he can.
0:34:48 But I’m not going to like make him do that.
0:34:51 And then we played with Replit and a few other tools.
0:34:54 And he just like loved the idea of like making stuff with AI.
0:34:55 He thought that was fascinating.
0:34:56 So I’m trying to do both.
0:34:58 I think most people should probably be doing that.
0:34:59 Totally.
0:35:01 We got to get your son on factory.
0:35:01 Yeah, I will.
0:35:01 I will.
0:35:02 Be careful.
0:35:04 He may like end up taking over your company one day.
0:35:05 Yeah.
0:35:06 Maybe I’d like a fun question before we go up here.
0:35:10 You know, so Matan, imagine you have a time machine, you know, you’re a physicist.
0:35:13 If I’m saying anything stupid here, just, you know, don’t.
0:35:16 You have a time machine, you travel to the year 2050.
0:35:21 Let’s say you get out in San Francisco, you know, and what do you see?
0:35:25 What’s different in the physical world, digital world, life, whatever.
0:35:26 Yeah.
0:35:30 I maybe have a hot take here, which is I think the arc of technology is actually exactly an
0:35:33 arc in that kind of where did humanity come from?
0:35:37 But like in nature, like hunter gatherer, there was no discernible technology other than
0:35:39 maybe some sticks and stones and tools and whatever.
0:35:44 I think we’re kind of about to hit the apex where like now you look out in San Francisco
0:35:46 and you see like Waymo’s, you see so much technology everywhere.
0:35:50 I think as time goes on, we’re going to kind of go back down and reduce the presence of
0:35:53 technology as much as possible, which you can kind of see the early starts up.
0:35:56 Are you saying everyone’s going to like destroy all the Waymo’s or what?
0:36:01 No, no, no, but just like people, as we get so much like efficiency and so much value
0:36:05 out of like medicine and technology and all this, I think you’re already seeing like
0:36:06 early inclinations of people trying.
0:36:12 there’s all these movements about, you know, cold showers or like MMA and all, or all this
0:36:14 like, you know, people spending a lot of time in nature or getting all this sunlight and kind
0:36:17 of going against the like pure technology for everything.
0:36:18 So the tech’s going to be in the background.
0:36:21 It’s going to be, it makes things amazing and it’s not going to be in our face.
0:36:21 Yeah.
0:36:25 And now I think this is very much SF is a bubble and the rest of the world is going to look
0:36:27 very different and we’ll adopt things in a slower pace.
0:36:31 I think San Francisco is going to kind of come back around and have a little bit of a
0:36:36 like, how can we have the same enablement with minimal presence from these things?
0:36:42 This also is skipping over the probably 20 years of like robotics that is going to be
0:36:47 at that apex, which I think maybe at some interim between where we are now in 2050, we’re going
0:36:52 to have by far a larger robot population in San Francisco than human.
0:36:53 I agree.
0:36:53 And I’m here in Kyoto.
0:36:58 I’ve actually been, you know, trying to advocate for the US and Japan to work together on this
0:36:58 stuff.
0:37:01 Because I think Japan would be a perfect place to be testing the robots too.
0:37:02 People are super open to it.
0:37:03 Oh, a hundred percent.
0:37:04 Yeah.
0:37:07 I mean, growing up, I was inspired by Gundam and all that, which, which came from.
0:37:09 I almost wore a Gundam shirt today because the whole droids thing.
0:37:10 That’s awesome.
0:37:11 Yeah, I almost did.
0:37:12 I agree with that.
0:37:15 And actually, I think, you know, OpenAI just announced the whole thing with Johnny Ive,
0:37:15 right?
0:37:16 Where they’re going to be building devices.
0:37:20 I was like, in one of the first episodes of the next wave, I said that I was like, I think
0:37:23 one of the big things in the future was like a future prediction is that the iPhone
0:37:25 is not going to be the last device for humanity.
0:37:27 It’s not going to be the last way that we interact with technology.
0:37:31 And the fact that we’re all just like staring down at our phones right now, like, you know,
0:37:32 looking like morons.
0:37:35 I’m hoping that eventually goes away and we have better ways to interact with technology.
0:37:37 And I think AI will actually enable that.
0:37:39 So it kind of fits with your vision for the future.
0:37:39 Yeah.
0:37:40 This has been awesome.
0:37:42 Is there anything you want to tell people?
0:37:43 Like, how can they get started with Factory today?
0:37:46 Yeah, go to factory.ai and get started.
0:37:48 We have 14-day free trials for everyone.
0:37:50 So go in, check it out.
0:37:54 If anything’s not up to your liking, or if you have any questions or thoughts, just shoot
0:37:55 me an email, matonatfactory.ai.
0:37:57 Happy to jump in.
0:37:57 Cool.
0:37:59 Is there anything special for Next Wave listeners?
0:38:00 Yeah.
0:38:04 So we have a very special deal for Next Wave listeners.
0:38:05 Awesome.
0:38:07 I believe it should just be in the link in the show notes.
0:38:08 Awesome.
0:38:08 That’s great.
0:38:11 Should people follow you on social media or maybe like follow Factory?
0:38:13 I think it’s Factory AI on Twitter.
0:38:13 Yeah.
0:38:15 Factory AI on Twitter.
0:38:16 Give us a follow.
0:38:18 We’ll be posting all of our updates, demos, that sort of thing.
0:38:19 Well, that’s awesome.
0:38:23 For anyone listening, you know, we’re trying to level up our game with this podcast and
0:38:25 hopefully, you know, you’re finding the episodes useful.
0:38:30 So please, if you would, it would mean a lot to me if you would subscribe on YouTube, you
0:38:33 know, if you’re listening on Apple or Spotify, subscribe there.
0:38:34 And yeah, thanks, man.
0:38:35 It’s been awesome to talk.
0:38:35 Thank you.
Episode 60: Can you really build an $8 billion SaaS startup by yourself using AI agents? Nathan Lands (https://x.com/NathanLands) sits down with Matan Grinberg (https://x.com/matansf), a physicist, AI founder, and creator of Factory AI—one of Silicon Valley’s best-kept secrets. Matan has published papers alongside luminaries and built a company trusted by top VCs and tech insiders.
In this episode, Nathan and Matan dive deep into the power and practicality of Factory AI—an agentic software platform that allows anyone to build full-featured SaaS applications using only natural language. After years of focusing on large enterprise clients and remaining under the radar, Factory AI is now opening up to everyone and revealing what’s possible when state-of-the-art “droids” (purpose-built AI agents) collaborate to automate the entire software development lifecycle. Watch them attempt to build a DocuSign competitor in minutes live on the show, and explore how AI is changing the future of engineering, entrepreneurship, and creative problem-solving.
Check out The Next Wave YouTube Channel if you want to see Matt and Nathan on screen: https://lnk.to/thenextwavepd
—
Show Notes:
-
(00:00) Enterprise-Focused Product Expansion
-
(05:45) Engineering Task Automation Tools
-
(07:01) Quick Project Setup Outline
-
(10:43) AI Revolutionizing Software Development
-
(14:29) Customer-Centric Problem Solving
-
(18:10) Progress Through Efficiency Improvements
-
(19:22) Agency: The New Success Metric
-
(24:54) Expanding Product to Small Teams
-
(25:38) Unified Platform for Software Development
-
(30:44) Importance of Foundational Knowledge
-
(33:55) Technology: Rise, Apex, and Decline
-
(35:40) Future Technology Beyond Smartphones
—
Mentions:
- Want the ultimate guide to use Gemini’s game-changing features? Get it here: https://clickhubspot.com/wdn
-
Promo link for 14 day free trial w 10M extra free tokens: LINK
-
Matt Grinberg: https://www.linkedin.com/in/matan-grinberg/
-
Factory: https://www.factory.ai/
-
Docusign: https://www.docusign.com/
-
Shaun Maguire: https://x.com/shaunmmaguire
-
Sequoia: https://www.sequoiacap.com/
-
Datadog: https://www.datadoghq.com/
-
Sentry: https://sentry.io/
-
Perplexity: https://www.perplexity.ai/
Get the guide to build your own Custom GPT: https://clickhubspot.com/tnw
—
Check Out Matt’s Stuff:
• Future Tools – https://futuretools.beehiiv.com/
• Blog – https://www.mattwolfe.com/
• YouTube- https://www.youtube.com/@mreflow
—
Check Out Nathan’s Stuff:
-
Newsletter: https://news.lore.com/
-
Blog – https://lore.com/
The Next Wave is a HubSpot Original Podcast // Brought to you by Hubspot Media // Production by Darren Clarke // Editing by Ezra Bakker Trupiano