YT YouTube2Text

YouTube2Text → Transcripts → YouTube transcript (8cSCJtPOkmM)

YouTube transcript (8cSCJtPOkmM) — Transcript

9,840 words · 1,547 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00Uh, so, I'm really excited about this
0:03talk. And can kind of fair warning, uh,
0:05the way I like to think about this talk
0:07is, um, like I sent it to Rob and Sonia
0:10ahead of time,
0:11is I said like this is about as close to
0:13like religion and politics as I think we
0:16get here. Uh, cuz there are some kind of
0:19personal beliefs and biases that I'm
0:21going to share, but I think this is the
0:24message that you all and I like need to
0:27hear. Uh, this is not the
0:30message that you might want, uh, but I
0:32think this is the thing that we all
0:34need. And this is not all good, uh, just
0:36to kind of preface it. Um, but but there
0:38is like the silver lining at the end.
0:41Uh, and so, in Micro MicroConf
0:44tradition, uh, this is my why. Uh, we're
0:47all here for a reason. This is my
0:49family. Uh, when when I very first
0:52started listening to the podcast and
0:54following Rob, uh, we were only three.
0:57Uh, and now we're four. And my son on
0:59the right here is taller than me. Uh,
1:01so, 13, 15-year-old kids and my wife
1:03Amanda. Um,
1:05this is why I do everything I do, right?
1:09Uh, to make as much money as possible
1:11and have as much freedom as possible so
1:13I can do things with them. Uh, and so, I
1:15think as we talk about this really
1:18sensitive subject, uh, I want you all to
1:20keep your why
1:22in mind because, uh, I think it helps us
1:25frame what's going on in our lives with
1:28AI and SaaS a little better.
1:35So, that's it. I told Rob this is going
1:37to be a quick talk.
1:40No. Uh,
1:42I I do think some of us are in
1:45big trouble because of AI. Uh, I think
1:48to ignore this, and this is the part
1:50that I think you need to hear, uh, is me
1:53not doing justice to y'all cuz I'm just
1:55like y'all. So, Castos, my SaaS
1:57business, been doing it for 8 years.
1:59Um
2:01I am worried a lot about what's going on
2:03with AI and the longevity of my
2:06business. And I wouldn't lie to you and
2:08say that I think it's all just like
2:09sunshine and rainbows, cuz cuz I think
2:11it's not. And if you assume that the
2:13next 8 years for me or 10 years for
2:16y'all or Where's Harry?
2:1820 years, right? Maybe? Yeah?
2:21Uh are the same, like I I think you're
2:23lying to yourself. And so, I'm going to
2:24walk through kind of why I think some
2:27things are different and will never be
2:28the same.
2:29Uh what we can all do to protect
2:31ourselves and our businesses and our
2:33whys against uh [snorts] the things that
2:36are coming and then how we can lean into
2:38and take advantage of these to be
2:40productive and proactive and and take
2:43advantage of the biggest technological
2:44shift, I think I've ever seen.
2:47Uh since the internet. Um because the
2:49fact is
2:51SaaS
2:52is greatly affected by AI. And it's not
2:55just SaaS, right? And it's not just tech
2:59that's disrupted by AI. And it's not
3:02just our jobs, like I think we all have
3:04jobs, right? But it's not our roles in
3:06our companies or jobs in general.
3:10I think it's the whole economy.
3:14And if you pay attention to the news,
3:16like
3:17this shouldn't be a surprise.
3:20So,
3:21this is about 2 months old, the
3:23beginning of the SaaS apocalypse.
3:26Uh this big drawdown, 285 billion
3:30dollars of equity erased from the
3:33markets.
3:34Anybody remember this? Why did this
3:36happen?
3:40This was as a result of a blog post by
3:43Anthropic about a bundle of skills in
3:45the legal space.
3:48So, a blog post erased whatever 300
3:51billion dollars
3:52in stock market valuation.
3:55Uh you know, whatever. You might think
3:57people in the public markets are idiots.
3:59I think sometimes, uh but but not like
4:01that much. They're not that wrong, I
4:03don't think. Uh and so, when things like
4:05this happen, I think we should pay
4:07attention and generally kind of be aware
4:09of what's going on.
4:12Uh this is a pretty cool-looking graph,
4:15I think, right? It's like uh what is it?
4:17Like uh what do you see when you see
4:20this graph? Like this is like one of
4:21those dreamscape things. Uh so, this is
4:23pretty cool graph, but again by
4:24Anthropic. And again, like they're
4:26really good marketers, right? Like we're
4:28going to talk about some of the stuff
4:29they're doing later and why you maybe
4:30shouldn't pay attention, uh but this is
4:33in the blue, potential disruption of
4:36jobs by AI, uh by job type, right? So,
4:39up top, like management, business
4:41finance, developers.
4:44Uh and then the red is like what's act-
4:46been actualized already.
4:49And so, if I come over here and see my
4:50notes, uh 94%
4:53of jobs could be replaced.
4:56Um
4:56and so far, like less than a third have
4:58been. And we've seen layoffs every week,
5:01right? Um
5:03and so,
5:04you know, this is from
5:07uh Andrej Karpathy, one of the
5:09co-founders of OpenAI. Uh all of the red
5:12here are jobs who will be negatively
5:16impacted by AI over the course of the
5:18next few years, uh with a score of like
5:21between one and 10, between seven and
5:22higher. Uh and this is $3.7 trillion
5:26of wages that could be lost.
5:29Um
5:30so, is this like everybody happy?
5:32Everybody like, "Yeah, this is [ __ ]
5:34great. Like I can't believe we're ending
5:36the conference on this."
5:37Uh there's just one more.
5:40Uh the iShares uh ETF that tracks like,
5:43you know, SaaS generally down 30% in the
5:46last 6 months.
5:50Told you this is like what I think we
5:52all need to hear. Um but like hey, we're
5:55entrepreneurs, we're founders, we've
5:57like seen this stuff before, right? Uh
6:00so yeah, you know, web frameworks,
6:03Laravel, Ruby on Rails, Next.js, Django
6:06made everything easier for us, right?
6:08Yeah, totally. Um cloud computing,
6:10right? I don't have to buy the $10,000
6:12server to start my thing. I can just go
6:14to Vercel and launch this thing for 10
6:17bucks. Totally true. Uh lowered the
6:20barrier, made it easier for us all to do
6:22stuff.
6:24Payments, right? This is uh Stripe's
6:26seven-line implementation of payments.
6:28Like again,
6:30uh made things easier for us all to do
6:32as entrepreneurs and founders. And I
6:34would argue a lot of AI is doing similar
6:37things, but I think it's also a little
6:39different. And so like all of these have
6:41enabled us to do more
6:43with less.
6:45And I think
6:47this time is different. And you might be
6:49like, yeah, it's different in ways, and
6:51maybe that's true.
6:53But I do think it's different because
6:55before
6:56all these advancements
6:58kind of lowered the barrier, made it
7:00easier for us to do stuff. But AI to me
7:03is just entirely replacing
7:06jobs and roles.
7:09And if I take like a half a step back, I
7:10think that's the big
7:13existential risk to us all is just like
7:17there's not going to be people with jobs
7:19in the medium term. And I'm going to
7:20talk about like the long term cuz I'm
7:21actually really bullish on the long
7:23term. But I think in the medium term we
7:25have a rough couple of years, not
7:27because of
7:28someone vibe coding our product. Like
7:30that's a medium kind of risk. I think
7:32the risk in the next couple of years is
7:34just going to be the economy.
7:36And there's not much we can do about it.
7:38But I want us to be ready to
7:42like prepare ourselves and our
7:43businesses to weather this time.
7:45And so, um
7:47Does anybody remember this MicroConf?
7:50No, this is a Gemini uh image, but I
7:52want to
7:53>> [laughter]
7:55>> But I want to uh I want to see, like,
7:58talking about replacing roles. Uh okay,
8:00so, i- if you've been in business for a
8:03couple of years, raise your hand if you
8:05were like hiring 2 years ago.
8:09Come on, raise them. Okay, cool. Uh so,
8:11like, half, two two-thirds of the room.
8:12Okay. Keep No, keep them up. Keep them
8:14up. Uh keep your hand raised if you were
8:16hiring a year ago.
8:20Cool, like most of us. Uh keep your hand
8:23raised if you're hiring today.
8:26Yeah.
8:29I don't Okay, that's a lot more than I
8:31thought, honestly. So, I'm like, oh,
8:32[ __ ] Um
8:35I talked to uh Liana mentioned, like, I
8:38serve as advisor in residence at
8:39TinySeed, where uh you know, companies
8:41coming in are between, like, a couple
8:43thousand dollars MRR to, like, gosh, I
8:45don't know, 50 grand.
8:47Um
8:49And they just got funding, right? So,
8:51they're like, I got to go I got to go
8:53spend this money, right? Cuz that's the
8:54whole trade-off. It's like, hey, we're
8:55going to give up a little equity, get a
8:56little bit of money so we can grow
8:57faster.
8:58Uh even within TinySeed, I don't talk to
9:02a lot of people who are hiring. Uh I
9:04certainly don't talk to a lot of
9:05TinySeed companies that are past like a
9:07year
9:09who are hiring. And I don't talk to a
9:11lot of
9:13bootstrapped SaaS businesses who are
9:15kind of established who are hiring right
9:17now. You know, the comment um from Oh,
9:19Victor's not here.
9:21Victor Keller, I've known Victor a long
9:22time. He's like, "Why would I hire
9:25anyone anymore because I can do
9:27everything I need to do
9:29with AI to run my business."
9:32Does anybody like resonate with that?
9:36Yeah, like almost as many people hands
9:38that went up originally. I think that's
9:40I think that's kind of the meta theme
9:41here is like cuz cuz I feel super
9:43empowered and like that's a good and bad
9:44thing sometimes, but I think that's like
9:46the immediate effect on kind of the
9:48economy and how we run our businesses.
9:53Um okay, so I want to talk about three
9:55and there's probably more, but there's
9:57kind of three that step out to me as
9:59like immediate kind of threats to my
10:01business and and probably yours. Uh and
10:03again, I say all this like I'm not like
10:05hopefully preaching, but like I think
10:07about this all the time cuz like again,
10:09my why, my family, like Castos pays our
10:11mortgage and our kids to play volleyball
10:13and stuff like that. And so I want to be
10:15really aware of this stuff.
10:16Um the first is probably the most
10:19obvious, right? Is that like our product
10:21just gets
10:22commoditized. And yeah, you might like
10:24roast me for saying someone can vibe
10:26code my product in a weekend. And like
10:27yeah, they're not going to vibe code
10:29Castos or Salesforce or whatever.
10:32But I think a lot of times we know like
10:34our customers don't need all of the
10:36things that we do, right? And could they
10:38do just like this one little thing to
10:40get this one reporter or this one way to
10:42store a file or this one API or
10:45something? Like yeah, like they probably
10:47could. And is that good enough
10:49to get on the other hand like the
10:51benefit of customization and like owning
10:54the data and and kind of having control
10:56over thing? Like heck yeah.
10:58Um
11:00I was talking to one of the Conversion
11:01Factory guys and I said like we Castos
11:04we charge $19 a month. So like first of
11:06all, don't ever [laughter]
11:08>> [gasps]
11:08>> start a SaaS business where you charge
11:09$19 a month. It is just it's rough.
11:12But uh
11:14ChatGPT's 20 bucks a month. Claude's 20
11:16bucks a month.
11:17For us to have the gall, I think in some
11:19ways, to say we are equivalent value to
11:23the most impressive piece of software
11:24ever,
11:26I I think is weird, right? Like I don't
11:28know like we we we churn hasn't really
11:30changed for us, but I think increasingly
11:33people are going like, "Gosh, I can do
11:34kind of like everything with Claude for
11:3620 bucks a month, and then I'm going to
11:38pay Castos 20 bucks a month." Uh I think
11:40that like the
11:42psychology
11:43is is kind of changing there, especially
11:45for like new customers. I think like
11:47we're starting to see some early signs
11:49of this and talking to with some of my
11:50peers, and they're like, "Yeah, like
11:51churn really hasn't changed, but like
11:53new trials are dropping."
11:55That that's that's kind of like the
11:56early sign of this, I think. And so like
11:58willingness to pay for a dedicated SaaS
12:01solution, and you see this on Twitter,
12:03right? It's like, "Why would I pay like
12:04500 bucks a month for a HubSpot when I
12:06can just make this myself?"
12:09Not true, right? But like not false,
12:11either, I think. Um you could use Notion
12:15and Claude code [snorts] and Airtable or
12:18whatever and get like a good CRM that
12:20does most of what you need from HubSpot
12:23for relatively free.
12:26So I think like the perception of the
12:27value of your product changing over time
12:30is absolutely a thing we need to pay
12:32attention to. And again, I think this is
12:34what we need to hear,
12:35uh even if you don't like it. Um
12:38I think the most interesting one is your
12:39team. I'm going to talk about what we do
12:42uh in empowering our team to lean into
12:44AI in a minute,
12:45but I think the most interesting thing
12:47here is us getting kind of like
12:49Is the word lobotomized where you just
12:51like go brain dead, right? I
12:54>> [laughter and gasps]
12:54>> I think uh I think this is a real risk,
12:56especially in kind of creative ventures.
12:58Like, "Hey, if all I can do is talk to
13:00Claude code or Codex or something like,
13:02what real value do I bring as a founder
13:05or a marketer or a developer, especially
13:08for kind of non-deterministic things
13:09like marketing, right? Creative things."
13:12Um
13:13And then I think there's this
13:14existential threat that we all are
13:17feeling right now of like, "What the
13:19heck am I doing here? If the robots can
13:22do everything, like what unique value do
13:24I bring? And what unique value does each
13:26person on my team bring when
13:28Yeah, not today. Like AI can't do
13:31everything, but gosh, like think about 6
13:33months ago. Think about 3 months ago
13:35what it was doing. Uh and then if we
13:37kind of extrapolate forward 2 years,
13:39like where are we going to sit in all
13:41this?
13:43And then I think this is the biggest
13:44one.
13:45This is the biggest one by far. Like I
13:47know there are products maybe here today
13:50that are basically a skill.
13:52Right?
13:54You're like a connector for this thing
13:55and you do this and you transform this
13:56stuff and then it pops out the other
13:58end.
13:59Like I think the reality is most people
14:01have no idea what Claude Code or Codex
14:03is and they don't know what skills are
14:05and so they haven't caught on yet. But
14:06if your business is a skill,
14:09the frontier models like just deliver
14:11this already. They have enormous
14:13distribution. Like literally everyone in
14:14the world uses them.
14:16And the functionality for this will be
14:18free.
14:20>> Hey, Rob Walling here. If you're
14:21watching this and thinking, I wish I
14:23could be in the room for talks like
14:25this. You can and you should. So much of
14:28the magic at our events happens in the
14:31hallway track. That's where you connect
14:32with two to 300 like-minded founders who
14:35are trying to solve many of the same
14:37problems you are. I hope you'll join us
14:39at our next event. Head to
14:40microconf.com/events
14:42to grab your ticket. Our next one is in
14:44Iceland in September of 2026 and then
14:46we'll be in Austin in April of 2027.
14:50>> So who was it that stole this slide
14:52earlier?
14:54I loved it. I was like motherf- I can't
14:56believe that that you asked this. But I
14:58think this is the Formspree. Was it
15:00Who's
15:01Yeah.
15:02This is so cool. I use Formspree. Uh and
15:05I I use Formspree exactly how uh
15:09this like is described. Uh it was
15:11suggested to me by Claude Code. I'd
15:13never heard Sorry, I'd never heard of
15:14Formspree before like 3 weeks ago when I
15:16converted my site over and it was like,
15:18"Hey, you need a way to send contact
15:20form and you should just use Formspree.
15:21Go here. They have a good free plan
15:22getting an API key and and give it to
15:24me.
15:27Right?
15:28Um so, this is a graph of Stack Overflow
15:30traffic over the last, whatever,
15:3320 years.
15:37That's not great.
15:39Um
15:40I mean, this is not surprising, right?
15:42Like, this was the way that many of you
15:44learned to develop is like, I'll just go
15:45to Stack Overflow, copy and paste stuff
15:47in and see how it works, and we'll kind
15:48of figure this out. Uh like, times
15:51change, right? This is not the most
15:53effective way to get value anymore. Uh
15:56and so, as a result, like, the market
15:57spoke, and they just get no traffic
15:59anymore.
16:01This one's super sad to me.
16:03Um Adam Wathan, founder of Tailwind CSS,
16:073 months ago, maybe, laid off 75% of
16:11their team.
16:12Very unfortunately, because the agents
16:14and the models knew the documentation of
16:16Tailwind CSS so well
16:19that no one had to go to their site.
16:21They had a 80% drop in traffic to their
16:23site, and I they sold, you know, premium
16:26products or whatever.
16:27But, the models and the agents were so
16:30good at implementing their free product
16:31that nobody had to upgrade to a paid
16:33plan.
16:35And like, Adam's a super smart guy.
16:36Like, I know him. Like, this not his
16:38fault at all. And this is none of our
16:40faults. Like, I think that's the one
16:41thing I want to say is like, what's
16:42happening in this transition of AI is
16:46not like my fault or your fault or any
16:49It's not Rob's fault even.
16:51>> [laughter]
16:53>> We didn't ask for this, you know? Like,
16:55the good and the bad. Like, the I
16:56definitely asked for the good, but we
16:57didn't ask for the bad, which is we all
17:00are forced to reinvent ourselves.
17:02Like, as of today. And I hope that's the
17:04one thing that you leave with is like,
17:06[ __ ] is going to get bad if you don't
17:07think about yourself and your business
17:09in a different way.
17:12Uh so, you know, moving to the good way,
17:14this is uh
17:15you know, the folks at Linear are
17:16amazing. Like, this is an image that
17:18they put out in their blog post about
17:20kind of the new version of Linear. And
17:22everybody in here uses Linear. Like we
17:24don't use Linear, but I know it's like
17:25the cool issue tracker. Uh, we just use
17:28GitHub. But like
17:29they went from like we're an issue
17:31tracker that replaces GitHub issues to
17:33like surrounding the entire ecosystem of
17:37agents, rules, context, and kind of that
17:40customer product feedback loop. Which
17:42like to me, as a non-technical founder,
17:45is the most important thing in SaaS,
17:47right? Like listen to your customers,
17:49make the [ __ ] they want, and deliver,
17:50and keep doing that over and over. And
17:52they're like pretty smart folks. And
17:54they're like, "Hey, why don't we just
17:55like say that we are the tool that lets
17:58you do that really well."
18:01So there's like an evolution, right? Of
18:03we're not an issue tracker, we're like
18:04the thing that lets you deliver on the
18:06most important function of a SaaS tool.
18:09SaaS business.
18:11So that's cool. Uh, like these guys are
18:13like the epitome of a MicroConf type
18:16business, right? Basecamp.
18:19Basecamp coming out with a CLI, right?
18:22This this is like where the puck is
18:25going, right?
18:26Is you have to have these kinds of tools
18:30to enable your customers to get value
18:33from your product the way they want to
18:35in the future. And like like Rob said,
18:37we are super early in this and very
18:39fairly we do not have a CLI or an MCP at
18:42Castos.
18:43Uh, partly because like we're not the
18:45kind of product that like has to have
18:47this today, I don't think. Anthony.
18:49Yeah, maybe maybe we're not. We we uh,
18:52we we debated it as of like 2 weeks ago.
18:54We think we have kind of bigger fish to
18:56fry here. Um, but we we considered
18:59putting off a very very valuable feature
19:03to do this. And it was really close. In
19:05the end we decided this is going to
19:06wait. Um,
19:09but I I this has to be part of the
19:11calculus one for discovery. Talking
19:13about Formspring, like, hey, how do
19:15people find us in
19:17increasingly? It's like, well, Claude
19:19Code said to use Formspring or Claude
19:21Code said to use Cast Docs or SavvyCal
19:23or whatever. Uh that's a that's a as,
19:27you know, to Amanda's talk yesterday,
19:29impressions maybe going up, but clicks
19:31are going down. I think part of that is
19:32like, we're just getting the answer from
19:35a different way. And if you're not
19:36showing up where that answer is
19:38generated,
19:39you're missing a big opportunity.
19:43Okay, so very early, right? When it
19:45comes to MCP and CLI and SDK and ABC,
19:49uh we're like really early with all this
19:51stuff, but I don't think that you can
19:53ignore
19:55where this is going. Cool?
19:57I promise this is going to get better.
19:59Um okay, so is it Britney? Like, what a
20:02girl wants? Is that Is that right? Yeah.
20:04What a girl wants. What does an agent
20:07want?
20:08What does an agent want?
20:12Context?
20:16Ooh, okay.
20:18Uh okay.
20:23Humans? Yeah, rent a human, right? Who
20:26would have thought we'd have a website
20:27that's like rent a human? That like
20:28isn't that really like wrong and illegal
20:31and uh
20:32okay, what what does an agent want?
20:34Anybody else?
20:36Energy, water. Yeah. Yep.
20:39Uh
20:42not a really pretty website,
20:45right? Like,
20:47uh I hate it for the designers out
20:48there, but uh they want an API, right?
20:51They want clean, visible documentation.
20:54It's just like, pro tip, uh
20:56a lot of our documentation is rendered
20:58on the page in JavaScript, and agents
21:00can't read that. So, if you're not like
21:03outputting your documentation as like
21:06whatever JSON or markdown or something
21:08like that. Like probably
21:09take the 5 minutes today to do that so
21:11that agents can see and index that.
21:14Uh an MCP CLI or SDK.
21:18And then like they want it free.
21:21They want it free and at most they want
21:23to like pay per outcome or something or
21:25pay per token or pay per usage.
21:27I think
21:29AI is
21:31overall incredibly deflationary.
21:34Uh we're all expecting more for less
21:36money.
21:37Uh we saw this with Ahrefs. Ahrefs
21:39historically is like [ __ ] gated
21:42everything in pricing tiers and overages
21:44and stuff and then they rolled back and
21:45they're like, "Hey, yeah, our API is
21:47like included in every plan now."
21:50They're kind of seeing the pressure of
21:51what we all expect in a uh
21:54in a like giving more away kind of mode
21:57and they're the market leader in their
21:58space.
22:00And so question for everybody like if an
22:02agent, you know, if my Open Claw tries
22:05to go use your product right now, like
22:07could it?
22:10>> Yes.
22:11>> Yeah, cool.
22:13That's the Formstack guy. That's
22:15awesome. Um
22:17Yeah, I mean for me for me the answer is
22:19no, but we're working on it. Um
22:22So this is like
22:24>> [laughter]
22:25>> this is like uh how I imagine myself,
22:27right? Like
22:29I'm like Dumbledore, right? I'm like a
22:33wizard who can run my company from the
22:35mountains in France or a boat in the
22:37Bahamas or whatever and we do provide
22:40value and it's like 24/7 always on.
22:42Anybody else like resonate with this a
22:44little bit? Like
22:45developers especially. Like I I this is
22:48like a developer to me is like, "God,
22:49you guys are magic." Like you guys can
22:52write code and develop products and all
22:54this kind of crazy stuff.
22:59Software developers aside, I think the
23:01the playing field has been leveled a
23:04lot.
23:05I I I I do some coaching and I work with
23:07someone here
23:09and and they sent me a message
23:11about a month ago
23:13and they go
23:15they they had kind of just learned cloud
23:16code
23:18and they were like,
23:19what the [ __ ] Like, if
23:21I can do this, everyone else can do
23:23this. And so, what what do I what do I
23:26do now? What value do I provide? How can
23:28I
23:29and they're they're an incredible
23:30founder and marketer. How do I
23:33continue to add the amount of value
23:36that I used to to my product when this
23:38is the new baseline?
23:43I don't know the answer. I don't know
23:44the answer, but I think that's kind of
23:45the question. And so, Alex Lieberman,
23:47right? Like, pretty smart guy.
23:49Anyone else
23:50>> [laughter]
23:50>> Anyone like, [ __ ] at 30 minutes I have
23:53an existential crisis the last 30
23:55minutes than the last 30 years. I think
23:58we're all kind of paralyzed, right?
23:59Like,
24:00what marketing channel do I go down?
24:02What product do I develop? Do I acquire
24:05a business or not, right? Like, I think
24:08I don't know what to do. I don't have
24:10the conviction to follow this course
24:14that like I used to. And again, maybe
24:15I'm
24:16maybe I'm projecting a little bit here,
24:18but I talked to enough founders that
24:20like they just they don't know. They
24:23don't feel really good about this is
24:25definitely the course.
24:26And so, if that resonates with you,
24:28like, you're not alone. But but I think
24:29we'll get into
24:30how to lean into something
24:32here in a minute.
24:33So, talk about motes yesterday a little
24:36bit. I think someone talked about motes
24:38and I was like, gosh, you guys keep This
24:39is the the curse of going last is like
24:41everybody talks about what you're
24:42talking about. I think there's a there's
24:45a couple motes. Maybe there's more than
24:46four, but I think there's four for sure
24:48that we especially as MicroConf founders
24:50and like mostly bootstrapped
24:53SaaS companies can really lean into
24:57the best is first-party proprietary
25:00data. So, if your product generates data
25:04that keeps getting better and more
25:06robust and more useful over time,
25:08you're good. Like, I don't worry about
25:10you, right? So,
25:11gosh, what's an example? Anybody have an
25:13example? Is this anybody?
25:16Yeah.
25:18What do you do?
25:18>> I do a ton of design research for my
25:20clients, so I gather and
25:23synthesize it among all of my other
25:25clients and
25:26make them all a bit stronger in the
25:28process.
25:28>> Cool. So, if anybody didn't hear, like,
25:30design research for a bunch of different
25:31clients and you're building this kind of
25:32library of, like, what great design
25:34looks like. Yeah, totally. So, you have
25:35first-party data a bunch of across a
25:37bunch of different experiences that you
25:38can apply to all your clients. That's
25:40cool. Anybody else? Example of
25:43first-party data that you're generating?
25:45Yeah, Keith, Segmetrics, right?
25:47>> We have data views and then we normalize
25:49data between multiple systems that's
25:51really difficult and AI is pretty bad.
25:54>> Yep. Yep. So, Segmetrics is
25:57uh attribution marketing attribution
25:58tool. Uh I mean, so you have I think you
26:01have uh proprietary data on two levels.
26:03One is like platform level, all of your
26:05customers you see a whole bunch of
26:07stuff, like, this is where people, like,
26:08you know, Stripe and ChartMogul can tell
26:10us how good or bad our businesses are.
26:11Like, that's really valuable. And you in
26:13a similar way. And then each of your
26:14clients have their first-party data that
26:18uh
26:19that's valuable to them, right? And that
26:20keeps getting better and more valuable
26:21over time the more they use your
26:22product. Yeah, that's really cool.
26:24Um this is where, like, uh we win,
26:26right? Castos wins here. We are a system
26:28of record. We're a podcast hosting
26:29platform.
26:30If you want a podcast, you got to have
26:32something like Castos. And if you cancel
26:34with us, your podcast goes away. So, if
26:37you're, you know, a CRM, uh as much as
26:40that's like a bad example, like, if
26:41you're the source of truth where, like,
26:44this is like this is where, like, our
26:47data for our company lives. I I think
26:49you're good. Um
26:52And because of this, I think these are
26:53harder to vibe
26:55>> [laughter]
26:55>> vibe code because like if you vibe code
26:58this and get this wrong, like the data
26:59that you think you have about your
27:00company is wrong and then it's not the
27:03source of truth, it's just a source of
27:04garbage.
27:05Uh and so
27:07that's one. Uh I think especially for
27:09this crowd, this is a super interesting
27:12one
27:13uh is speed.
27:16I'm much more bullish about the people
27:18in this room with SaaS and AI. I am
27:23I don't want to keep using HubSpot as an
27:25example, but like a legacy big
27:28frankly bloated uh piece of software,
27:31right? They just can't move as fast as
27:33you can to AF AI-ify yourself and your
27:38team and your product and your company.
27:41And so
27:43being able to move fast, not having to
27:45report your decisions to a whole bunch
27:46of people and keeping that product
27:49customer feedback loop really really
27:50super tight is maybe the universal
27:54advantage that everyone in this room has
27:55that you should really lean into.
27:59And then like yeah, hey, what if my next
28:01customer is an agent not a human? Well,
28:03like that that's one thing, but the
28:05other thing is like if you're selling to
28:07the enterprise, which like increasingly
28:08I think that's where SaaS should go. Uh
28:11again, like don't start a SaaS recharge
28:1320 bucks a month. Um
28:16if you're charging thousands of dollars
28:18a month, your ACV's 50 grand or
28:20something like that.
28:21Uh people always have bought from people
28:24and they will always buy from people and
28:26maybe even more so this relationship
28:28between you and someone else is
28:30paramount. You know, agents aren't going
28:32to decide to buy a $50,000 a year piece
28:36of software. It's going to be you and so
28:38like leaning into sales
28:41uh I think is a really interesting
28:44avenue to continue to succeed
28:48in SAS.
28:51Okay, so I want to like another show of
28:53hands. Like how many of you feel good
28:56about how your team is using AI?
29:01Oh.
29:03Cool.
29:04Okay, a little of this, right?
29:06Uh and not like really high. Okay, cool.
29:08Um
29:11I don't do a lot of things right, but
29:13but like we [ __ ] nailed this. Like
29:15our team is cracked, super clawed
29:17pilled. They're amazing. And so I want
29:19to share what we're doing. Um
29:22Okay, so
29:23uh and I I think like to take a step
29:25back,
29:27this has to start with you.
29:29It can't be like, "Oh, our developer is
29:32going to, you know, have their
29:33development team like use all these
29:35codes and develop all these best
29:36practices." Like even if you're not a
29:38technical founder, which like I'm not,
29:41you have to lead by example. Like I I
29:44always want to be the dumbest person in
29:46the room in our company. And a lot of
29:48times like I am, but in this respect, I
29:52led the charge. And I think you have to.
29:54You have to set the culture of
29:57excellence here. And like we're not
29:59messing around. Like this is how it's
30:00going. And I'm going to talk about some
30:02really challenging times we had.
30:05But you have to get this right and y'all
30:07have to lead this. It's no You can't be
30:09like, "Okay, I'm going to appoint like a
30:10head of AI in our company." Even if you
30:13do that, that person's not going to take
30:14you seriously if you yourself aren't
30:17fluent and proficient at these tools.
30:19Okay?
30:20So,
30:23uh in our company, we're seven people
30:25full-time.
30:26Everyone has Claude code.
30:28Support,
30:30product, marketing,
30:32certainly development. Everyone uses it
30:34all the time.
30:36Every We have a team meeting every
30:38Monday, all hands, all hands all seven
30:40hands. Uh
30:41every time it's like, "Hey, I was
30:43playing around with this new MCP over
30:45the weekend and we like built this new
30:47tool."
30:48"Hey, I was looking at this API for this
30:50other thing we could do. It's really
30:52cool. Like, I built this script that we
30:54can run to get this report."
30:57And it's because
31:00that's the expectation. I led by example
31:02for a painful
31:04>> [laughter]
31:04>> amount of time
31:06being the person that y'all make fun of.
31:07Like, "Oh my gosh, Craig's going to talk
31:09about AI again and how great it is." And
31:10all my god, let's just roll our eyes. I
31:12did that for long enough and we had some
31:14positive peer pressure in the room to
31:16where they were like, "Oh, I probably
31:18should like catch up."
31:20Or like, "I'm going to get fired." And
31:22that was the conversation a couple
31:23times. Um
31:26We have an unlimited budget for AI.
31:28Um
31:29I I can't imagine telling someone that
31:32you pay 100 grand a year that they can't
31:35upgrade their cloud plan from $100 to
31:37$200 a month. I just can't imagine that,
31:39right? And we're going to talk about
31:41what the kind of bigger companies are
31:42doing around this in a minute.
31:45Um we have the expectation, not the
31:47allowance, but the expectation that
31:49everyone spend at least 2 hours a week
31:52learning. And they're like, "Well, where
31:53would I learn?" I'm like, "You can watch
31:54my YouTube channel."
31:55>> [laughter]
31:56>> Because it's frankly like really good.
31:59But if you don't watch mine, like watch
32:00somebody's
32:01um
32:02around like the area where you want to
32:05like learn. Uh because
32:07you have to keep learning cuz this stuff
32:09changes so fast. I was like just before
32:10this, like they're going to release
32:12uh Opus 4.7 like tomorrow and then
32:15OpenAI has GPT 5.4 security, which is
32:19kind of their answer to Mythos, I guess.
32:20But like, literally it changes all the
32:22time. I was talking to someone about
32:23Paperclip and I was like, "Gosh, like
32:25I'm in this all day every day and I have
32:26a hard time keeping up." But if you as a
32:29leader don't
32:30uh set the expectation with your teams
32:33that they have to keep up with this
32:34stuff, there's no way that they will.
32:36And so this has to start at the top and
32:37then come down.
32:39Uh so, so we have two standing meetings
32:41in our company. One is like our Monday
32:43all-hands, and then we do this thing
32:45called co-working, which is just like
32:47hanging out on Thursday. But, hanging
32:49out turned into what we did this morning
32:51in the lobby, which was tell me what you
32:53learned or show me what you did with AI
32:55this week.
32:57And this does something really
32:58interesting. So, the first time this
33:00happens, you got a couple of people who
33:01show up and stuff. And then I call on
33:03someone. I'm like, "Hey, Jimmy, what did
33:05you do this week?"
33:08Oh, I don't know. Like, not much.
33:11Jimmy, next week that cannot be the
33:13answer.
33:14Like, you got to show up. You have to
33:15show me what you've learned and how you
33:17are furthering yourself to be able to do
33:19your job better in the future. And and
33:22like
33:24this just has to happen.
33:26Um
33:28and I think this is
33:30where um you guys know the the term like
33:33right person in the right seat?
33:36Yeah. Uh what's the traction guy? Um
33:41I was having this conversation with one
33:42of you this week that like you're like,
33:44"Hey, I'm super on board. Like, I'm
33:46Claude pilled, but uh my my development
33:49team is not. And I'm having to explain
33:51stuff to them." And this was a
33:52non-technical founder.
33:54Person was like, "Did Do I have like a
33:56people problem here?" I was like, "Fuck
33:58yeah, you have a people problem. Like,
34:00that is not okay. And and we had
34:04several of these conversations with our
34:06team. And the the last sentence or the
34:09kind of the the question that you always
34:11should ask yourself in hiring. I got
34:13this from Ruben Gomez. Ruben's like one
34:14of the smartest founders I know. Follow
34:16him on Twitter. Uh is knowing what I
34:18know today, would I hire this person
34:20again?
34:23So, I ask myself that of everyone on my
34:25team as it relates to AI. Knowing what I
34:27know now, would I hire this person again
34:30in the kind of AI world that we live in.
34:32And the answer is about 50/50.
34:35Yes and no.
34:36And so, I just told them that. Like, it
34:39doesn't have to be any more complicated
34:41than that. Hey,
34:43Jimmy, I'll use Jimmy as an example.
34:44He's like the most amazing
34:47human, uh and he started off like, "This
34:50is just not I'm not going to do this."
34:52And I was like, "Bro, like you got to
34:53get on board. Like, I love you, but like
34:55for to work here, you have to
34:56participate like this or you can't work
34:58here anymore." And since he's the
35:00biggest transformation in our company.
35:02And so, if y'all have those folks who
35:04are like, "Oh, I'm too good to use cloud
35:07code to write code cuz I'm an artist on
35:10the keyboard."
35:11Like, that person has no place in my
35:13business and probably shouldn't in
35:15yours, either.
35:17And so, right person in the right seat
35:20doing the right job
35:22had a meaning 6 months ago and I think
35:24has a different meaning and context
35:26today. And so, y'all as leaders might
35:31need to have challenging conversations
35:32with your team.
35:37Right on. Thanks, man.
35:39I thought y'all were asleep. Um okay,
35:41cool. So,
35:42200 companies in Tiny Seed, 200 founders
35:44in Tiny Seed. Uh get to talk to a lot of
35:47them about what they're doing.
35:49And you know, just as a as a side note,
35:50I have 9 minutes left. Like, we're going
35:51to go over, I think. I hope that's okay.
35:53We'll
35:54we'll still get to the bar. Um but, we
35:56might go over a bit. Um
35:58It's the beauty of going last. Uh
36:01so, what other Tiny Seed companies are
36:02doing? This is a bit rapid-fire. So, uh
36:05hired an AI native engineering manager.
36:07I like this a lot cuz this is like,
36:08"Hey, we had this job before of like
36:10engineering manager, really kind of
36:12tough job, I think. Now, it's like, how
36:14can I pill my developers as much as
36:17possible and get them just token maxing
36:20as much as possible, right? This is this
36:22person's job. I love it, right? We had
36:24this other job before, now we have the
36:26new version of this job. I think there's
36:27a lot of that going on.
36:29Uh
36:30automated like task audits, right? If
36:32something happens more than a couple
36:34times, we're going to automate this.
36:35Because now everyone is empowered to do
36:38so much around like I'm going to build a
36:40script or I'm going to build a little
36:41tool or I'm going to automate this thing
36:42or I'm going to get this thing in Claude
36:44code or open claw or whatever, right?
36:46This should be happening
36:48uh
36:49to everyone, not just your developers,
36:52right? Because your developers are super
36:53good at like, "Oh, if I have to do this
36:54thing, I'm going to build a bash script
36:55to like automate it in the future." Now
36:57everyone can do this, which is a huge
36:59unlock cuz then it's multiplying
37:01exponentially.
37:02Is that a thing, multiplying? It's
37:04happening exponentially.
37:05Um
37:06token consumption.
37:09Brandon, here, right? Uh the expectation
37:13is your engineer should be consuming 60%
37:15of their token quota in a week or month.
37:18Uh yeah, like, "Bro, what are you
37:20doing?" Like, I wonder sometimes like,
37:22"What are my developers doing if they're
37:24not maxing out their $200 a month plan?"
37:26Cuz I can max mine out and I spend like
37:28a couple hours a day futzing around with
37:30this stuff.
37:31Uh what are they doing to not be asking
37:33for like a $1,000 a month
37:36quota? I don't know. They should be
37:37maxing out all the time.
37:39Uh this is kind of similar to our our
37:42weekly uh kind of show and tell,
37:442-minute demo of a thing that you did.
37:46This can be like a skill or an
37:48open-source thing that you did or a way
37:49that you're doing like Codex and uh
37:53and Claude code like review or it can be
37:55whatever, right? Like how this happens
37:57in your business can be different, but
37:59you should be doing it. This positive
38:00peer pressure is hugely
38:03uh transformative, I think, in
38:05instilling the culture of kind of like
38:07AI excellence.
38:10Uh
38:10I like this kind of like
38:11cross-functional squads. Like we've kind
38:13of always done this and now it just has
38:15a AI flavor, right? So like pick a
38:16thing, have a couple people just go
38:18super hard on this for a period of time
38:22uh and then share what they did with
38:23everyone else.
38:26Gosh, this one's pretty tough. We
38:28definitely don't have this nailed.
38:31The way that we do this most of the time
38:32is we just publish our stuff open
38:34source. And so everyone can kind of keep
38:36keep track of it. But like how do you
38:37share prompts and workflows and skills
38:40amongst your company
38:42in a way that's like consistent and not
38:44like getting in the way. And I know like
38:46if you're on the Anthropic team plan,
38:48you can share skills and stuff like
38:49that. But not everybody is in that boat.
38:50So like have a way to do this
38:52however you achieve it, I think is
38:53important.
38:55Talking about token maxing,
38:57evidently at Meta, they have like a
38:59whole leaderboard around this, right?
39:01Like how can I spend the most tokens
39:04possible?
39:05Um
39:07This is where like again, y'all y'all
39:09roast me for being so like into this
39:11stuff. This is directionally right, I
39:14think. Like yeah, do you want to like
39:16use the most tokens possible? Like maybe
39:18that's not exactly like proper, but do
39:21you want this culture of like
39:23I did this thing with AI and it's
39:25awesome and I want to show you all about
39:27it. Like that's definitely right to me.
39:29Um
39:31Jensen Huang, founder of or CEO of
39:33Nvidia,
39:34we definitely don't have $500,000 a year
39:37>> [laughter]
39:38>> developers. But I think the the gist
39:40here is right, right? If your developers
39:43aren't maxing out their plans and asking
39:45for more, you got a problem. And I would
39:47take this step further, not just
39:49developers, but like if you have a sales
39:51person or a marketing person or an ops
39:53person and you haven't given them Claude
39:55code yet and they're not banging down
39:57your door,
40:00I I would fire them, like literally
40:02today. Cuz it's like where what are you
40:04doing? Like if you're working for a SaaS
40:06company
40:07>> [snorts]
40:07>> and you're kind of plugged into even if
40:08it's LinkedIn and you don't know about
40:10this stuff and you haven't tried, you
40:12know, Claude code worker or whatever and
40:14they're not asking you for more access
40:16to tools like this,
40:18that's a personal problem. And to me,
40:20that's like
40:21an irrecoverable personal problem.
40:23That's a person that just doesn't care.
40:26To to me.
40:32Um
40:33I think a talk on AI would be incomplete
40:35without talk of Open Claw. Anybody using
40:39Open Claw in their business?
40:42Not a lot of hands. Yeah, cool.
40:44I want to talk to y'all afterwards. Um
40:48Open Claw, everybody know what Open Claw
40:49is?
40:50Yeah, cool. Um so, I I think the thing
40:53that Open Claw and now Hermes, which is
40:55kind of my
40:56uh tool of choice here,
40:58did is take us for
41:00take us from I'm at my computer doing a
41:02thing and I have AI that's really cool,
41:04uh Claude Code or Codex, right? To this
41:07thing is always on, it's proactive, it
41:10can work on a schedule, it does things
41:12with inputs that are not me.
41:14So, Peter Steinberger, right, sold Open
41:17Claw to Sam Altman and Open AI for a
41:21billion dollars probably, right? Like
41:22pretty cool, first billion-dollar
41:23company, a single-person billion-dollar
41:25company probably.
41:26Uh
41:28not an amazing uh sort of an amazing
41:29product, but much more of an amazing
41:31mindset shift, I think, for all of us,
41:34right? Like Claude Code, still my daily
41:36driver and the place I derive most value
41:38when it comes to AI.
41:40Uh but, if you fast forward
41:433 months now, you see Anthropic adding
41:47little bits of what Open Claw is to
41:49their suite of products.
41:52Right? Remote control, dispatch, right?
41:55All these kinds of things. Uh and sure
41:57enough, like the final form that I think
41:59this will take is the Codex app or the
42:03Claude app doing things that Open Claw
42:05did.
42:10So, [snorts]
42:12again,
42:14uh Jensen saying every company needs an
42:16open claw strategy or an agentic
42:18strategy.
42:21Does everybody feel like they have a
42:22strategy around this? Like
42:24for customers or for your team to be
42:27empowered with agents?
42:31Sweet. Awesome. One. I I don't. Two.
42:34Cool. Awesome.
42:36Yeah?
42:37Awesome. Awesome. I like uh
42:41We're getting there. We're getting
42:42there. Um I want to share something that
42:44that like I do, but this is one place
42:47like we've not rolled this out to our
42:48team. But if you think about like
42:50amplification of someone's skills, like
42:52surely figuring this out for your team
42:54members,
42:56not just your customers, is the next
42:58like huge leverage point. Um so, the
43:01place where I'm I'm kind of doing this
43:02right now is for myself. So, I call this
43:04my chief of staff. I hope the QR code
43:06works. If not, let me know. Uh
43:09but this is a chief of staff
43:12set of skills and set up for Hermes. So,
43:14Hermes was like the previous slide,
43:16uh
43:17is a thing like open claw, but it
43:19doesn't suck to me, right? It doesn't
43:21break. It's way more secure. It's run by
43:24uh
43:24a research lab. So, they're like kind
43:26of, you know, not just vibe code Well,
43:28they're probably vibe coding, but like
43:30they're doing it better. Uh the thing I
43:31like about Hermes better than open claw
43:33is the memory. It's amazing. It it it
43:36learns by itself and has kind of
43:38self-improving memory over time. It's
43:41awesome. Uh so, my chief of staff here
43:43is like what a chief of staff should do,
43:45right? It clears my calendar. It checks
43:48and kind of categorizes my inbox.
43:51Uh and like then you build on top of it.
43:52And so, that's kind of the goal of this
43:54repo is like, "Hey,
43:56this is a starting point. Get you two
43:58steps down the road and then build as
44:00you will from here."
44:04Is this like resonant Oh man, I got a
44:05minute left. Sorry, we're going over. Um
44:08Does this like Does this seem like
44:10anyone's company today?
44:13This is my company like a lot of ways
44:14this is my company. Like you got you got
44:16amplitude over here. You got search
44:18console over here. You got analytics in
44:20this place and that place and none of
44:21them talk to each other. And I think
44:22that like if you're going to take one
44:24thing away from this
44:26it should be this concept. So like this
44:29is
44:30life today because this is kind of like
44:31as good as we could do, right? We have
44:33all these disparate tools in a bunch of
44:34different places
44:36and like they don't talk.
44:38And that's challenging. But that's like
44:40gosh, we're so fortunate to have all
44:43these amazing tools, many of which are
44:45free, right? Or cheap.
44:47But like
44:49this is the vision.
44:51This is like where you want your company
44:53to go to be AI powered. I think where
44:56you want your company and your product
44:58to go to be AI powered where you have
45:00all of the inputs, right? All of the
45:03data about your customer and where they
45:04come from and what they're doing in your
45:06product and how they're interacting with
45:08it and getting value
45:10and then you do AI stuff with that
45:13information and you deliver a better
45:16value to your customer on the other
45:17side.
45:19And honestly, that's
45:22where we are in our company is we're
45:24ripping out kind of everything about how
45:27data flows and how we relate that data
45:29and value to our customers
45:31and putting it in a place where we can
45:34AI-ify
45:36everything that we do for our customers.
45:39Because you think about like here's a
45:41really specific example. It's just like
45:42your website, right? So your website's
45:44in
45:45WordPress, right? And you're like,
45:47"Gosh, I'm going to write a blog post
45:48over here in Cloud Code and I'm going to
45:50copy and paste it into WordPress and
45:51then I'm going to go to Google Analytics
45:53and see how that did and then I got to
45:54look at search console and then I got to
45:56look at like Stripe to see like how many
45:58people converted and stuff and like
46:00if you can do all that and like make
46:02sense of it, that's amazing. But what if
46:04like
46:05all of that was in one Claude Code
46:08project and you just be like, "Hey, tell
46:09me about my customer acquisition from
46:11the last 30 days and how we can improve
46:13it."
46:16And it would do a really good job, I
46:17think. Like you might need to write it
46:19to SQL Lite or something like that.
46:21But this is where again like that
46:23product customer feedback loop allows
46:25for rapid uh iteration and testing and
46:30better measurement, I think, for
46:32everything that you do in your company.
46:33And so something like this probably
46:35needs to be where most of us are going
46:38to interact with
46:40the things that we create for our
46:42customers in a better way.
46:46Okay, so um
46:49this is where the talk gets better. Uh
46:51I'm actually really really really really
46:54bullish on the future. Uh there's an
46:56asterisk to that, but I'm really bullish
46:59on this. Um
47:00I think that open claw
47:04uh this is uh is it hyperbolic? Changed
47:06the world. It it changed how we think
47:08about what AI can be from a I sit at my
47:11computer and chat GPT and like ping-pong
47:13mentally back and forth to like
47:16it it this doesn't happen, right? It's
47:17like Napster, right? Like Napster was
47:19not the final version of what music
47:22sharing is,
47:23uh but it changed how we think about it.
47:25And I think open claw is the same,
47:26right? I think
47:28agents doing work and providing value
47:30for us to our customers is absolutely
47:32happening and will continue to happen
47:34and I'm super bullish about it. And this
47:36is the thing that tipped the scales and
47:38started us going in that direction.
47:42Uh
47:43I think agent to agent
47:45commerce
47:47absolutely is happening and will
47:48continue to happen in the future.
47:50Uh and
47:52you should probably be one side of this
47:54conversation, right? Uh
47:57is SaaS dead?
48:01Kind of.
48:04Kind of. I think even Rob kind of
48:05alluded to this, right? Like and your
48:07Q&A's like gosh, like I'm I'm not like
48:10this isn't really SaaS. It's like a an
48:12API tool. Like some some SaaS's are kind
48:14of just API tools, but like
48:16is the value we're going to deliver to
48:18our customers just agentic and just
48:21programmatic in the future and there's
48:22no interface? Like yeah, I think so in
48:26some circumstances. And so if you're not
48:28thinking like how can I be the guy on
48:30the left here to where other agents like
48:32open claw and whatever is going to come
48:34in the future uh like I that's where I
48:37worry about myself and and like you
48:39might should too um because I think that
48:41like
48:42the way we deliver value through
48:44software
48:46will be through agents.
48:48Or whatever form agents take in the
48:50future, but it's probably not an
48:51interface where people go to a website
48:53and type things into a form.
48:56Um
48:57vertical
48:59way over horizontal. That's pretty easy.
49:01I think the reason here is this is just
49:03so much more defensible, right? You have
49:05subject matter expertise in a way that
49:08makes it harder for other people to
49:11replicate the value that you deliver.
49:15Uh open weight models, open source
49:18models. Anyone running open weight
49:19models, open source models? Cool.
49:22Awesome.
49:23Um
49:24this has gotten a lot better like in the
49:26last
49:28what? 3 weeks, right? Quinn 3.5 and 3.6,
49:32Gemma 4. Uh they're really good. Like I
49:35would say gent like y'all again roast me
49:37afterwards. Gemma 4 is like Haiku right
49:40now.
49:42Really good. If you're just like, "Hey,
49:44I'm going on vacation to to Italy this
49:47summer, give me a 5-day itinerary."
49:50Way plenty good. Uh 90% of the things
49:5390% of people do, you can do with open
49:56weight models today. And so, as y'all
49:58are thinking about like, I'm going to
50:00AI-ify my company, but gosh, like every
50:02time I introduce AI to my product, like
50:05my costs and my COGS increase. Well, not
50:08necessarily, right? Get you some GPU or
50:11run this on a a server somewhere.
50:13Zero cost, 100% data privacy.
50:16Right? If you have
50:17Harris talk, you know, people saying
50:19like, "I don't want my data going
50:20anywhere." Cool, like local finally is
50:24an option here in things like Turbo
50:25Quant from Google, which like compresses
50:28the memory needed by some of these
50:29models by six times,
50:31uh really kind of opens the playing
50:32field for this to be
50:34uh possible.
50:38I'm getting a robot. I don't know about
50:40y'all. Like, [laughter] I'm about a year
50:42away from doing the dishes and laundry
50:44and like mowing the grass.
50:46Uh
50:46this is what I tell my kids, right? This
50:48is what I tell my kids.
50:50Uh they both are very kind of like
50:52I know I'm over time. I'm sorry. I'm
50:54almost done. Uh robotics is 100% where
50:57the puck is going, right? AI is going to
50:59leave the computer and going into things
51:01that we do, and this is where I get
51:03super excited uh about like the future.
51:05And so,
51:06if we want to take a step back and like
51:08get some perspective here,
51:10uh I think like it's not
51:13as negative as maybe I made it out to
51:16be, but I don't hear any wallets like
51:18gut checking us. So, I want to really
51:19like give us some perspective here.
51:22Um I absolutely think the future will be
51:26amazing, right? It's going to be like
51:28Wakanda, right? Like everything is going
51:30to be
51:32free, basically. We're going to have
51:36food and medicine and resources and
51:38infrastructure in ways that like we
51:40can't imagine today, right? Cuz imagine
51:42like
51:43there's been a one-person billion-dollar
51:45company probably already. Like that was
51:47unthinkable six months ago. And so with
51:49the pace of change that all of this is
51:51happening, 100% like we are all going to
51:54be better off
51:55in a few years. There's no doubt in my
51:58mind. And and it's those few years that
52:00are the challenge, right? It's these
52:02next couple of years where we're going
52:04to get through the transition of like
52:07the conventional economy to like where
52:10things are going and the the panacea, I
52:13think that we're all going to be in. And
52:15so we have to survive, right? And I got
52:17to talk about like project glass wing
52:19because like I think this is
52:22where this could all fall apart for us,
52:24right? Is like we all run software
52:26companies and things like the Mythos
52:28model that, you know, supposedly are
52:31going to like
52:33you know, find all these exploits in our
52:35software could could be like very bad
52:38for us as software companies. Uh and so
52:41I think that like we have to stay aware
52:44of like the negative potential when it
52:47comes to bad actors using AI to do
52:49really bad things uh
52:51with our companies.
52:54And so like uh I'll wrap it up here. Uh
52:56so
52:57MicroConf founders have kind of three
53:00things that we should be doing. One,
53:01move really fast, right? Shipping it
53:03testing all the time.
53:05Uh two, like stay in your lane and niche
53:07down as much as you can, I think.
53:10And then like leverage the human-ness
53:13and relationships that you have and can
53:15have as founders, right? Getting on
53:17calls as the founder with your customers
53:19is something that no one else can do.
53:22And so if y'all are like this is cool, I
53:24want to like keep having this
53:25conversation. I'm doing a three-day,
53:28call it a challenge, like where I talk
53:30in more depth about what I'm doing at my
53:32company to get us ready for the next few
53:35years. It's free. I have absolutely
53:37nothing to sell you. Uh but you can sign
53:39up. We're going to do it on Zoom over
53:40your three days next week. It'd be cool
53:42to have you all there. I can share a
53:43little bit more about what I'm doing and
53:44give us kind of frameworks for how you
53:45can set your company and yourselves up
53:47to win over the next couple of years.
53:51>> [applause]
53:54>> I've been running MicroConf for more
53:55than a decade, and the thing I hear from
53:57attendees more than anything else is I
53:59wish I'd come sooner. Not because of any
54:01single talk or tactic, but because they
54:03finally found their people. Founders who
54:06get it. Founders who are building the
54:08same way they are. At our events, it's
54:10not just about the talks. In Portland,
54:12we organized waterfall hikes, food
54:14tours, happy hours, real time with other
54:17founders outside the sessions. And those
54:19relationships don't end when the event
54:21does. They turn into the people you text
54:23when something breaks or when something
54:25finally works. If you've been thinking
54:27about coming, I hope you'll join us in
54:28Iceland and Austin. Get your tickets at
54:30microconf.com/events.
54:33If you got something out of that talk,
54:35make sure you hit subscribe. We're
54:36putting out more of these every week.
54:39And that wasn't the only great talk from
54:40Portland. Go watch this one next.

About this transcript

This page contains the full transcript of YouTube transcript (8cSCJtPOkmM) , generated from the public captions YouTube serves with the video. The transcript has 9,840 words across 1,547 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.