Subscribe
Half Baked
Sep 17, 2026
4 AI Business Ideas Too Good To Ignore
4 AI Business Ideas Too Good To Ignore
00:00
51:43
Transcript
0:00
I wrote a whole plan. I started building a prototype. I wanted to raise an incredible seed round. I got all my investors lined up, of which were some very, very well-known people.
0:09
The day before I took all their money, I decided to kill the idea. You've started companies for 30-plus years, had acquisitions from Google, BlackBerry. Founders think their idea is unique. It's not.
0:21
Over the last 10 years, we've tested north of 500 ideas. We just created our 45th company. Timing-wise, this is just opening up a brand-new opportunity and industry.
0:31
There's a non-zero chance that Claude just wins everything, but the average end user is not going to think through this combination of- Yeah, hopefully you get Patrick Carlson on your, your cap table, and if you did, I think you'd be in very, very good shape.
0:43
The America's Cup before he built a brand. Sold it to BlackBerry, then another to Google. Deal signed, that's McCannon, startup studio veteran. And now back at it again.
1:01
Well, TA, welcome to the oven. I'm hot and ready to be here. [laughs] I love it. I think of all the guests we could possibly have on the podcast, TA, you probably fit our hand like the best glove anyone could ever buy.
1:19
You've started companies for 30-plus years. You said you built your first website back in 1995 before anybody had an email, and, you know, you've had acquisitions from giants like Google, BlackBerry, RIP.
1:35
Uh, you've been with Pioneer Square Labs for, you know, eight years.
1:39
You've probably seen and saw, been involved with more companies than John and I could ever dream of in 10 lifetimes, uh, alongside some pretty good outcomes there.
1:49
But the big thing that I am so excited about here is that I've never met someone working on a company currently with the exact same mission as Half Baked, to create a million founders.
2:01
And to make it one step better, I'm 83% Irish. 83? 83% says 23andMe. 83% Irish. [laughs] I love it. Does that mean some of the accent, though, is gonna come out? I, I, gosh, I hope so. I don't really have the accent.
2:18
It's a little bit far removed from the late 1800s, but, uh, but it's still in the genes. Do you know where in Ireland? Uh, my grandmother was from southern Ireland, County Cork. Oh.
2:29
And my grandfather was from Tempo in Northern Ireland. John, tell TA why that is music to our ears. Well, you're talking to the, you know, two biggest, uh, Cork podcasters in the world [laughs] right now.
2:43
Uh, that's actually probably not true. Uh, yeah, no, both myself and Daragh, uh, we hail from, uh, from, from County Cork as well, so I was laughing. Uh, you know, 23andMe, uh, says you're 83% Irish.
2:54
You're probably more Irish than I am. I've actually never done a 23andMe test, but I'd actually, I wouldn't be surprised if you're more Irish than me.
3:00
So, uh, yeah, that's, uh, definitely something in common we have, which is, uh, which is really cool. I'll get, I'll get the pronunciation incorrect, but I think she was from Loughbarron, B-A-R-R-H-A-N, out on the coast.
3:15
God, I've never even heard of that. No, I've never even heard of that. Yeah, so don't worry. No one's, no one's gonna be critiquing your pronunciation for a place we haven't even heard of, TA. So- [laughs]...
3:25
let's just say, let's just say that's the right one. Um- So- I think, like, d- you know, this podcast is all about ideas. You've come to us with
3:34
so many ideas, uh, that, that we're gonna dive into, but probably makes sense to do a bit of context setting here. Yeah, great. Well, let me just do a q- just the quickest bit of, um, sort of background.
3:44
So I'm a mechanical engineer, so I've always been kind of a person who likes to build machines to build other machines, um, and startups are some of those.
3:52
Um, I really started building companies, I mean, I actually built some as a kid, but then built my first sort of professional companies, you know, when the internet was happening, and now I feel like I've built companies through every big technology wave.
4:05
Uh, after I sold my fifth company to Google, I joined Pioneer Square Labs, and that is a startup studio or an incubator. Uh, over the last 10 years, we've tested north of 500 ideas.
4:15
We just created our 45th company, and we've built a whole machine to turn ideas into companies or validate or invalidate most ideas, so that's what we do all day long at PSL.
4:26
And as the engineer here, I've been sort of always the guy who's like, like, "How do we add more software and add more process to what we do so we get repeatability, scalability?"
4:36
And we basically did one version of software, another version of software. Uh, GPT-3.5 came out. We basically added AI into it.
4:46
Reasoning models came out, and we started showing this software that we had built for ourselves to other studios, other founders, other incubators, other accelerators, and they all said like, "Man, I want some of that.
4:56
Like, I want to use that product." And so we rebuilt it one more time, and that's become Lev. So Lev is an AI co-founder.
5:05
It basically takes this process of ideation, validation, and early operation and productizes it for all the things that founders should be doing.
5:14
So, you know, in some ways, Lev is 30 years of my experience and 10 years' worth of PSL's experience all put in a box and made available to founders everywhere, and in the same way that you're trying to do, we're hoping that Lev can help many more founders thrive, uh, in a way that can help them focus on the best ideas, the right ideas, uh, and, you know, really democratize entrepreneurship, you know, around the globe.
5:40
Great pitch. [laughs] I feel like after hearing all there is to know about Lev, you're gonna come in here with some steamroller ideas- [laughs]... AI-driven.
5:52
Uh, before we actually get into the ideas, TA, you know, the sentence alone around, you know, validating, working on 500-plus ideas-
6:01
I mean, surely there are some trends and patterns within tho-that, that validation and the signals that translate across, you know, many different ideas.
6:09
What are some of those kind of instinctual learnings that you've had that you can apply to ideas that you have now? Yeah.
6:16
Well, w-at PSL, you know, we are specifically trying to build venture scale ideas, and those are specific kinds of ideas. They're not bootstrap companies. They're not restaurants down the street.
6:26
They are specific kinds of ideas, and that's not to say they're better or worse, but when you're doing a venture scale idea, it has a certain set of criteria to it.
6:35
Uh, the ways that we've sort of productized that at PSL is, you know, we, we are going to move through as many different ideas as we can to ultimately get good ones out the door.
6:45
Some things that we do both in terms of PSL, but then also as productized in Lev is, is first of all, is a lot around competitive landscape. You know, most founders think their idea is unique. It's not.
6:57
You know, there's usually five guys working on the same idea at that point in time. You just didn't know about it.
7:02
And so one of the first things that we really do, and many of our ideas die on competitive landscape very quickly. You know, we find five other people who've already been doing it for two years.
7:10
They already raised money from the best investor, dead. So we kill very quickly on that. The second is really around customer discovery, customer development.
7:21
You know, how quickly can we characterize an ideal customer profile? How quickly can we go find somewhere between 10 and 30 ideal customers to go talk to?
7:30
And lots of times we scream out into the internet and, like, no one gives a shit. And like, okay, dead, or we go and talk to customers, and they have all different types of sort of problem solution fits.
7:42
So while the founders, like their happy ears, are hearing the same thing, they're actually not. So we kill a lot of times on what I would call customer development, customer discovery.
7:51
We just can't find enough people who care about the exact same thing that want a solution to it. The next is ability and willingness to pay.
7:59
Like, most founders don't want to ask that question because most people don't want to pay for stuff.
8:04
So the-- when we find that we, you know, are asking those customers like, "If I built this thing, would you pay me X amount of dollars?" So we kill a lot on ability and willingness to pay.
8:14
If we get through those three big gauntlets, then we will get into somewhere of is it feasible to build? But let's say that most cases, most things are buildable.
8:25
And then the next piece would be around market size, or the last piece would be around investor interest. There are certain categories that are very exciting for investors, and there's lots of categories that aren't.
8:35
And so if we find any one of those combinations of things in almost that order, um, you know, those are kills, reasons we kill. And, you know, we, we celebrate kills here at PSL just as much as we celebrate successes.
8:48
And on every Monday when we go through and we're like, "Well, we killed that idea, we killed that idea, we killed that idea," everyone claps because that gives us an opportunity to move on to the next idea.
8:59
I've got to ask, though, is there any idea that you guys have killed that l- in retrospect, you definitely shouldn't have killed? Have you had an-any kind of example of, of, of that?
9:08
Well, I think that there are plenty of examples of companies that we went too slow on. We had the right insight, and we just didn't execute fast enough.
9:16
So part of that is we didn't execute the product fast enough, and more importantly is we always want to start-- We have to start a company with a founding CEO.
9:24
So a lot of ideas we have here that are tracking really well, but we just can't find a CEO in the right amount of time. Uh, so I mean, we had, we had an idea for what effectively has become Devon way before Devon.
9:37
Uh, and, you know, we had a product that was working very well as a software development agent. We just executed it too slowly, and that has turned out to be a very important, you know, company and companies.
9:49
Uh, and you know, there's a bunch of those ideas that we either executed slowly, we couldn't find the right CEO for, um, and you know, there's got almost lots and lots of them that are like- Let's dive into some ideas you have 'cause there's a few here that I definitely want to double click on.
10:05
I think one in particular is this protocol idea. So, um, we share a little bit of history here too, is we're, we're both really, you know, athletes.
10:15
So I went to college on a swimming scholarship and was, you know, highly optimized, went to Olympic trials in swimming in 1988, and then I quit, uh, I finished my swimming, started, uh, you know, being a robotics engineer, and then I quit to become a professional sailboat racer.
10:31
So I spent- [chuckles]... about seven years as a sailboat racer and optimizing in a different sport.
10:36
And so I've always been a sport optimization person, and now I would say I'm a, you know, uh, trying to optimize on health, longevity, human performance.
10:45
So I've built multiple versions of how do I collect all my data, how do I understand myself well, and how do I, quote unquote, optimize for either human performance or health longevity, productivity.
10:59
So, uh, you know, the protocol idea is really around leveraging all this incredible power that lives in these LLMs, specifically Claude, to build a personal health record, and that personal health record can be applied for, uh, any given situation.
11:16
So for example, I've torn my meniscus in my ri-right knee. I'm tracking all different kinds of things in my right knee.
11:21
I'm thinking about all the different ways in which I could improve upon that, whether that be physical therapy or some kinds of newfangled, uh, interventions.
11:31
Uh, I'm also tracking on any kind of, you know, health condition I have. But it's been amazing to me, and I've been prototyping this along the way of how do I effectively help myself have Dr.
11:43
Claude, you know, als- alongside me all the time, and that requires both a tight harness.
11:49
It requires some level of how am I collecting and organizing the data and the power that sort of lives inside Claude specifically, but it could be an optimized LLM or a, uh, or a different LLM, uh, to really improve health and human performance.
12:06
It could be even more uniquely applied When you think about people who have unique health conditions,
12:14
oftentimes those unique health conditions are difficult to track, uh, they are difficult to communicate, and they're difficult to find correlations.
12:23
So the LLMs, and I've found my own experience, like correlation and causation has been incredibly useful to find, uh, in this sort of protocol product that I've been kind of playing around with on the side, you know, for myself.
12:37
Um, last but not least is for people who have critical health conditions that span multiple disciplines, so these are oftentimes like autoimmune diseases, where you go to a doctor and they're like, "You know, I just don't understand, like, this thing that is happening to you," and you go to one doctor, the next doctor, the next doctor, the next doctor.
12:57
You spend two-thirds of your time bringing doctor number two up to date on what does doctor number one tell you.
13:03
So in my prototypes of Protocol, the ability to characterize for a generic next doctor in the list or to characterize for a specialty or subspecialty is all there.
13:15
It's very, it's incredible, like, how this could all work, really helping people take control of their personal health with a goal of health or human performance, uh, depending on the person. So that's Protocol.
13:28
I think it's only, it's only a matter of time before something like this happens, because timing is so good. I mean, if you think about it, like, there's been two bottlenecks.
13:35
The bottlenecks is collecting the data, and now, you know, they're, you know, the cost of sequencing a genome has gone from millions of dollars to a couple of hundreds of dollars. Yep.
13:44
You can now get blood work done really quickly. You have these wearables that's collecting data on your HRV, your heart rate, your stresses, does blood pressure,
13:54
and it's converging nicely with the ability to actually interpret that data. LLMs coming in, actually being able to take this massive amount of information and dial it down to a couple of insights that's actually useful.
14:05
I'm conscious this is quite a big idea, a big problem. Like, there's a b- there's a big difference in my mind between human performance and health.
14:14
So, because health would be very much around, like, how you might be able to, you know, uh, influence what you take for drugs and, like, pharmaceuticals and, like, making sure you're taking the right products, whereas performance is more about, you know, how do I make sure that my ACL heals really, really quickly or that the rehab is good?
14:29
How, how would you think about, you know, narrowing it down to a, "Okay, I have this concept, I have this idea, I have this problem. Timing is right." What do you do next?
14:38
Well, if I were doing it myself, um, then, like, uh, you know, advice I would give to anyone is, like, you gotta pick your ICP, right?
14:46
And in this case, I would probably pick, I would probably not pick the human performance side. I would probably pick the critical health condi- what I would call critical health condition side.
14:56
And again, like some, one of these autoimmune diseases or a, kind of a weird disease that somebody has that they're struggling with, kind of capturing enough information to find the insight and to share that insight with other specialists, and those other specialists could be, you know, ones that they know or ones that they don't know.
15:14
Uh, and so I think I would probably pick that, but, you know, that's where you get back into founder mentality as, you know, certain founders will care about the, "Hey, I wanna go to the human performance side and go to elite athletes," uh, and you go to the Tadi Pogacha sort of realm.
15:29
Or you're gonna go for somebody who has struggled with a particular thing and can't quite find a solution for it. You mentioned as well that you're prototyping the idea at the moment. What does that look like in terms of
15:44
what kind of, uh, you know, what stack are you using to actually build out a prototype that you can use kind of yourself? Yep. So this is a Claude project.
15:52
I'm happy to share, you know, anybody who wants to kind of see where I, how far I've gotten on it, uh, specifically.
15:58
Um, and I've uploaded as much data as about myself as I possibly could, so this has got microbiome stuff. This is, uh, genetic stuff. This is blood tests, et cetera.
16:10
Uh, I personally was struggling with an iron deficiency, and I couldn't figure it out. Doctors couldn't figure it out, and basically Claude figured it out. So I had a weird, uh, a weird, uh, bacteria in my microbiome.
16:25
I have a genetic variation that makes it difficult for me to a- to absorb iron, and I was taking my iron supplements at the wrong time. Incredible. Just incredible. If you proposed this idea 23 years ago,
16:40
it would've cost you $2.7 billion to sequence your genome. [laughs] Yeah. Uh, in 2006, they brought that down to 300,000.
16:51
By 2007, it was a million, 2014, 1,000, and now it's 200, some claiming they can do it for $100 or less. Like, that's insane.
17:01
Ti- timing-wise, this is just opening up a brand-new opportunity and, and industry that just wasn't open yet.
17:08
The one thing that I'd love to get your take on, TA, when it comes to this idea is, you know, you have lots of people that are able to collect the data, so you have, you know, WHOOP to collect your, like, the sense data day to day.
17:21
You might have Function Health that does the blood work. You might have some other provider that might sequence your genome and loads of other things that I can't remember from my, you know, 15-year-old biology class.
17:31
How do you think this idea layers, you know, in tandem with all of them?
17:35
Because, like, my sense is that, you know, you can get involved in the actual gathering of the data game, uh, or you can get involved with the orchestration layer on top, where you use the data as inputs, and it's your job to then do something with it.
17:52
Yeah.
17:53
I mean, this is, if we're thinking about it as just a, you know, a fun project or we're thinking about it as a business and an idea, then again, back to the sort of the, you gotta start to choose, and you're gonna- Mm...
18:04
build a con- a consumer service like this, or are you gonna sell it into, you know, somebody else as a, a support for a concierge medicine type of thing? I think I would probably go in the consumer direction.
18:17
Now the problem you get in with consumers, like everyone freaks out that you're gonna basically be getting medical adv- advice from some, you know, untrusted LLM type of thing, and that, that's a totally reasonable [chuckles] issue.
18:28
That's a thing that a founder's gonna have to go solve for. Uh, but there's lots of, you know, do- lots of ways to improve upon that.
18:35
So in terms of the, um, what I have found in my own exploration is you can very quickly get outside the bounds of something like a function health in a very positive way.
18:48
So for example, if you go to function health and say like, "Hey, I've got this torn meniscus in my knee," they're not gonna do, they're not gonna do anything for you.
18:55
Or are you gonna say like, "Well, could you actually start giving me some exercises for it? And now can I report when I do those exercises, my knee swo- w- was swollen, but it didn't hurt very much.
19:05
Oh, it's hurting more on the inside, not the outside. Oh, when I did this hiking and I did this and I used my poles, it was better, but it wasn't." You know, that one way or another.
19:13
So this opportunity to go from effectively a baseline understanding of my health, of which my genetic background doesn't really have much to do with my torn meniscus, but it might have lots to do with an iron deficiency.
19:27
And, you know, the, the productization or corporatization of one of those ideas is gonna draw the bind- the bounds relatively narrow, like a function health or, you know, a primary care doctor or something like that.
19:38
So, you know, Protocol is interesting in that way, and it could be easily built out in a couple different directions if you're talking about just personal health or you're talking about a subs- a sub-area like a known disease state, where you're looking for other people who have similarities in that known disease state.
19:56
You're looking for what interventions. This is where I would b- bridge it into an idea I call N of one. So if you think about experimentation, like all of us are unique until we're not.
20:07
If you had the same torn meniscus and I did at the same point, if we could find each other and you could say, "Hey, I've been doing this exercise, you should try that exercise," or, "I've been using this supplementation," or, "I tried PRP, you should try PRP."
20:20
This opportunity to build community around a particular s- specialty that is built around data sharing and search and discovery. So this is like a related but different idea I call N of one.
20:31
John, this does bring up something though that I remember you talking about before. You had an idea that you wrote in the newsletter on a tweet that Patrick Collison had.
20:41
Yeah, so this was about three months ago, so I'm obviously constantly scouring Twitter for, for startup ideas. That's basically my, my full-time job.
20:48
Um, and yeah, ultimately, um, yeah, a couple of months ago, Patrick, uh, was tweeting about, uh, exactly that. So, um, I actually have it here. I can, I can read it out, uh, exactly what he said.
21:00
So he said that, uh, "By far the most u- by far the most usef- useful preventative medical device that I've ever received has come from unleashing coding agents on my genome, having them investigate my specific mutations, and having them recommend specific follow-on tests and treatments."
21:14
So, like, when I read that, my immediate thought was, like, that's obviously, like, a really good startup idea.
21:19
Like, ultimately, you know, when you think about, you know, what makes the good basis for a good startup idea, if you're able to, uh, you know, ride these kind of mega trends, let's say, um, that's always a good jumping off point to, uh, to kind of start a business.
21:33
So obviously you're talking about that kind of intersection of like AI and specifically AI agents, which is obviously the biggest, you know, mega trend in the world right now.
21:41
You've then also got longevity on the other side, which is kind of like, I would say, nearly the second biggest, you know, trend at the moment.
21:46
Um, uh, the analysis costs something like $100 to do, um, with the kind of AI agents. Sequencing the genome costs a few hundred dollars.
21:55
And yeah, essentially, like, I think it's one of those things where how do you, how do you democratize that and how do you get that out into the world?
22:02
Um, so the idea being like how do you create a kind of pre- preventative medicine platform which basically uses AI to analyze your genome, investigate what mutations you have, and then obviously recommend specific, you know, treatments and, uh, things that you could do.
22:15
Yeah, hopefully you get Patrick Collison on your, your cap table, and if you did, I think you'd be in very, very good shape going forward. [chuckles] Patrick, call me. Call me, I'm listening.
22:23
[laughs] If Patrick Collison is listening to this, I think-- I, I hope he has better things to do with his time than listen to this podcast. I really, really do.
22:31
[laughs] Patrick, if you're, if you're listening, please call me too. That'd be great. [laughs] Well, this is, this is one of those, one of those kind of ideas, you know, hopefully,
22:44
hopefully most founders can answer the question, which is, "If I'm successful and I cre- create a really big business, how's the world a better place?" I hope that most founders ask themselves that question.
22:55
Now, you can't always have the best answer to it, depending on the particular idea. But this kind of idea and many ideas is sort of like, if this worked, like the world would be a way better place, and that's exciting.
23:08
Yeah.
23:08
It's, it's exciting to think when you think about startups like that, like what you're doing, what Lev is doing, hopefully what Protocol will do or somebody will do Protocol, and maybe we're a small influence on what they're doing or maybe we can invest in what they're doing.
23:22
But the, the world would be a lot better place if these people who have critical health conditions or want to live at human performance levels could understand what is the right intervention, what is the right strategy to ultimately improve.
23:36
Yeah, Jeff Bezos said something similar.
23:38
He was talking about how, um, if he does his job right, uh, his impact on the world through the business he starts and, and funds will dwarf anything he can do in terms of, you know, charitable contributions and, and stuff like that.
23:50
Um, which is obviously still great to do, but he's talking about how like, yeah, like entrepreneurship is such a, an incredible force for good in the world when you have, I think, the right people working on the right problems, as you said.
24:00
So interesting. Uh, I'll, I'll send it and you maybe include it in the show notes, but I almost started this company in 2017. Hmm. It was a whole personal health record company. I had... This is a good founder lesson.
24:12
So I had a very similar idea prior to LLMs, like prior to what we can do today. I wrote a whole plan. I started building a prototype. I wanted to raise an incredible seed round.
24:23
I got all my investors lined up, of which were some very, very well-known people. Um, the day before I took all their money, I decided to kill the idea. Why? It just wasn't good enough at the time, and I'll tell you why.
24:39
And, um, I wrote a long blog post about my, my logic on doing that.
24:44
But the lead investor at that point, he said, he said, you know, when I went, I flew to see him, and I told him the story about why I was gonna kill the idea. And he said, "You know, I liked you a lot before.
24:53
I like you even more now." But, you know- [laughs]... making difficult choices is part of the founder journey. And so I, I almost started a personal health record company called LifeWell 2017.
25:03
Uh, it just wasn't the right time.
25:07
At the time, the majority of your health data lived, at least in the United States, it may be true around the world, it lived in a personal health re- it lived, lived in an, uh, electronic medical record owned by your hospital, and your hospital had no real interest in you having that data.
25:22
And it was very, very difficult to get that data, and they were effectively averse to any one of us taking control of our own health.
25:30
And they controlled the data, and they controlled the access to the data in such a way I didn't feel like I could get around the misalignment of incentives, and so I ultimately killed the idea for that reason.
25:43
But now- It-... partially because of laws, partially because of data access, partially 'cause you can collect a lot of this data on your own, you don't need your hospital anymore. You don't need- Mm...
25:52
your doctor anymore.
25:53
You can go do these tests independently, and in many ways, many of these tests can be done in such a way that they're more important, and you can build your parallel health record, which is a superset, significant superset of what you might have at your, you know, your primary healthcare pro- provider.
26:11
If you're listening to this podcast, there's a good chance you have an idea that you want to bring to life. And whenever anyone brings an idea to me and asks, "What do I do next?"
26:20
I always say the same thing, "Make it real. Give it a home."
26:23
And for all of the businesses that I've ever started, alongside the thousands of businesses that we have helped start through Half Baked, that home is Google Workspace.
26:31
It gives you a business email address, so people actually take your idea seriously.
26:36
You get one place for all your docs, your calendar, meetings, storage, spreadsheets, [laughs] and even AI, with Gemini built right into it.
26:44
Setting all of that up used to take me days, and now it takes less than 10 minutes. Nothing else even comes close on price or quality, and anyone who says otherwise, I challenge you to do.
26:53
Google Workspace are very good friends of Half Baked, and they're giving our listeners and readers 14 days free and 20% off their first year. And with plans starting as low as $7 a month, you really do have no excuses.
27:06
So if you want to grab this exclusive offer, go to gethalfbaked.com/google. That's gethalfbaked.com/google. Thank you, Google Workspace, for making this episode possible. What else have you got, TA?
27:24
What else have you got cooking in the oven? Uh, I want to talk about Scout. So- Let's do it...
27:30
uh, I'm a, I'm a, I'm a productivity geek, and I've used, uh, executive assistants, uh, offshore executive assistants for a long, long time.
27:40
Uh, my wife went out of town, and I said, "You know, I'm gonna try to rebuild all the executive assistant stuff in Claude." And slowly but surely, I have built a full team.
27:50
So Scout is my EA or my COO, and Scout now has, I think, 13 specialists that do different work for me. Each one of those specialists has a specific job that they do.
28:00
For example, one that's called Mainstage, which prote- prevent or sort of, um, preps me for podcasts. One that deals with all my different boards and board preps. I just went through an acquisition.
28:11
I have another one called Quorum that helps me through acquisitions, uh, and so on and so forth.
28:16
And so now I have a good process by which I have Scout, and Scout builds specialists, and Scout manages the specialists, and now Scout can even distribute itself to other people.
28:27
So anybody who wants a version of Scout can email me after this, and I'll send them one, uh, and we'll see how we go with some level of viral distribution of effectively your own EA, COO.
28:40
And it's pretty well-focused today on specific things. My, let's call it a business user, and like I have a travel planner, I have a gift manager, I have things like that.
28:53
I have a daily brief manager, and then some sub-spec- subspecialties around the specific work that I do, like making investments and sitting on boards and things like that.
29:03
But I've been-- I've had a really fun time building Scout, and now Scout is, you know, a critical part of my workflow, and I think I've used Scout today... Let's see. I did my daily brief. I prepped for this podcast.
29:16
I had two board meetings. I have three board summaries to do, uh, later on today, and I have a presentation I have to do tomorrow, all of which Scout is managing for me while we talk.
29:26
How involved are you in the whole loop? Um, I... A little bit, and a little bit in the sense of like, like, "Scout, give me d- my daily brief." "Oh, I have this podcast." "Okay, Mainstage, go do that."
29:41
"Oh, I have a birthday coming up. Go to Keepsake." So Scout is managing a majority of that, and I'm, I would say doing primarily input.
29:51
So I sit down at the end of my day, Scout runs a end-of-day workflow, and then I just talk back at, in this case, Claude/Scout, and I just say like, "Oh, I finished that thing for Dara. I need to do that thing for John.
30:02
Oh, that's a P1 thing. Oh, this is a P3 thing." I just talk at it, and effectively then it just does its work for the next day.
30:11
Is it fair to say that like, when I think about both Protocol and Scout, the thing that kind of, the through line that's, say, between two, the two ideas is that they're all about like personal context.
30:20
So in your opinion, is that where you kind of see things going in terms of, you know, there's a lot of talk about where does the value accrue in the AI stack, you know?
30:29
Uh, going forward, do you think that a lot of that will accrue in the ideas that kind of play on, you know, having, let's say, context on, on a person, whether it's their personal life, their, their health, their business?
30:39
Um, is that how you see things playing out? I guess I would say there's, there's a non-zero chance that Claude just wins everything, right? There's a non-zero- [laughs]
30:48
Like, but I- They're, they're winning a lot of things right now. [laughs] I mean, there's a non-zero chance it's just, like, it, it is the application layer, and it just knows. However, I think there is something to
31:00
person- let's say, context and carrying context, whether it's Protocol or it's Scout or it's Lev. Those are all forms of context. Two is there is value in design. So how is the information presented?
31:15
So we could all imagine what a daily brief looks like, but I think my daily brief is really, really well designed to get a business person to be highly productive in a day.
31:25
So there is value in design, just like there's value in Protocol design, or there'd be value in Lev's design. And then I would say the last piece is the workflow.
31:35
So what is the workflow that you think about for, like, how would we use Protocol? There's, like, lots of different ways to think about the workflow. Mm.
31:44
And but the average end user is not going to think through this combination of context, workflow, and design. Mm-hmm. And then somewhere in there, again, I mean, these guys down at the Anthropic are pretty smart.
31:58
Like, they will think about this, but, but you l- if you look at CoWork, like, there's little bits of this starting to happen in CoWork. But the... You know, is CoWork gonna go build Protocol? I don't think so.
32:11
Is CoWork gonna go build Lev? Probably not, because you need the context of the creator, in that case, you know, us- Mm...
32:18
but then also the context as it's organized on behalf of the end user, and then presented in some way that also has a workflow, right?
32:27
And the workflow could be an alert, it could be a request, it could be a script, it could be a cron job. But the, those I think are, are important components to any one of these sort of application layered things.
32:41
I think, I think the hope is that, like, AI, uh, ends up being a little bit like Y- what YouTube did for the creator economy.
32:48
So if you think about, you know, pre-YouTube, if you wanted to create, you know, content, you had to, you know, have a show. You had to have a production team. It would cost millions of dollars to make something.
32:57
Whereas, you know, YouTube came along.
32:59
All of a sudden, if you had, you know, a webcam or whatever, you could, you could make YouTube videos, and you could blow up and, and get them in front of millions of people, um, or, or a very small number of people if that's what you wanted to do.
33:10
I think that software is probably gonna go the same way, that you're gonna have, like, way, way, way more entrepreneurs, you know, enabled by AI, by Claude, by, by things like that.
33:19
And ultimately, they'll be able to build for the kind of long tail of, let's say, companies or users that it wouldn't have been feasible to build for previously because it was just too expensive to ship software fundamentally.
33:30
So I think that's kind of where I like to think. That's where the, the optimist in me thinks it's going, and yeah, the pessimist thinks that just buy Anthropic stock and, and, and give up on everything else.
33:39
[laughs] That's the, that's the other way of thinking about it. Don't say that, John Rowe. We're out of business if people do that. Yeah, that's a good point. That's very, very true.
33:46
[laughs] I, I agree with you, John, I mean, in the sense that the opportunity to think about, you know, think about and conceive of ideas which may be not this big and s- but this big,
34:00
and be able to execute that big very, very rapidly. And again, if it's, if you're, if it's ephemeral, meaning it lasted for a month, three months, six months, 12 months, 24 months, who knows?
34:12
But there are, there will be value delivered in those periods of time that will just be subsumed at some point or, or not. You know, and every now and again- Mm...
34:21
you'll come up with the right idea that isn't competed by somebody else that does turn into one of those gigantic ideas. But both of those are okay outcomes. Both of those are okay outcomes.
34:32
Whenever I go running, I always come up with my best ideas.
34:36
For example, the other day when I went for a run, I thought about the idea of wearing glasses that had augmented reality built into them that could make my runs a bit more entertaining.
34:46
And though I definitely don't have the capacity to go and start this idea now, maybe this is gonna go in the someday pile.
34:53
And any ideas that I always feel a little bit more conviction on than others and go into the someday pile, I always buy the domain, because you never know when you're gonna use it, and the best time to buy a domain for a business that's gonna be launched in the future is yesterday.
35:09
And the place that I always go to buy my domains is name.com. They are a longtime friend of Half Baked.
35:14
In fact, they were the second sponsor we ever had back when we were a fledgling media company, and they have been supporting us ever since.
35:21
So if you have some ideas in your someday pile or even your today pile, go and buy them from name.com. Thanks, name.com for keeping the lights on here at Half Baked.
35:31
Tia, you mentioned, you know, like flexing the muscle about getting from, uh, idea to value as quickly as possible.
35:38
Like, for Scout and with all of the different agents that it has, how would you go about going from the idea that you currently have right now, which you're using yourself, to a go or no go for this as a business case?
35:53
Yeah. Well, the first decision I would say is you have to decide whether you're trying to do a venture scale company or a bootstrap company, right? Or even this ephemeral company.
36:02
And let's c- let's say as you move along that continuum, an ephemeral company is almost like it's just for fun. [laughs] It's like, "Hey, if we made beer and pizza money, that's fine," right?
36:11
This is more for fun and interest or more for personal value, like paint- doing a painting or playing guitar. And I personally get a lot of joy out of the creation process of something like Scout or Protocol or whatever.
36:25
So, so you could say it's like, "Oh, who cares if it's a business? It's something that gives me pleasure." In the creative process.
36:32
Then you move into bootstrappy thing like, "Oh, I can, I hope I can make beer and pizza money out of it."
36:36
But if you make this choice that you're going down the venture path, there's a whole different set of choices there, right?
36:42
Both in terms of your commitment and how many people you're gonna try and get with you, and how big it has to be, and how much better it has to be than anybody else.
36:48
So, you know, that's, you know, that's an important decision around, you know, any, any particular idea. What would you say,
36:58
like, if I am an 18-year-old and I'm coming out of college, and I'm ready, I'm ready to start something, I think there's this, you know, dichotomy of people saying, "Hey, you should buy a laundromat and get cash flow," and then you have other people saying, "Hey, you should go and create the next Stripe or Air- Airbnb."
37:17
You know, when it comes to, like, self-assessing on whether, you know, you have an idea like Scout and you're unsure whether to go one path or the other, how would you kind of go about deciding that if you were an 18-year-old coming out of college?
37:33
Yeah, my, my advice to most young people is try to go... If you think you want to be an entrepreneur at some point, go start working alongside another entrepreneur who's s- who's skilled.
37:44
Like, go get your job, go get a job at Stripe. Go get a job- [laughs]... at Spotify or PSL or whatever. Go try to work alongside somebody who's already... And it doesn't have to be a big-scaled company.
37:55
It could be a well-regarded, smart, experienced entrepreneur.
38:00
That's point one, because you're gonna start to learn things, and you're gonna be around other people who've been attracted by those people that you just wouldn't learn otherwise. That's point one.
38:09
Point two is start doing side projects. Like, start doing half-baked. Like, "Hey, last week I worked on this thing. Oh, I threw it away. I worked on another one, threw it away. I worked on a third one, I threw it away."
38:20
'Cause you're learning about the process of entrepreneurship, and oftentimes you're gonna start learning about other people you want to work with, and, like, why do you like working with John versus Dara, or the other way around?
38:31
And that process for the near term, meaning you still have to build companies with other people, uh, and I think most people will still want to build companies with other people.
38:42
So, start doing side projects with the goal of that being a learning experience and a creative endeavor, and every now and again you'll come up with this particular idea. And do not quit your day job until you just can't
38:57
spend any more time on the side project.
38:59
Meaning, like, slowly but surely, you'll start doing a little side project, a little bit more, a little bit more, a little bit more, a little bit more, and at some point you'll, you'll have enough momentum.
39:08
And a good founder can effectively do two jobs at the same time.
39:13
They can do a full-time day job, and they can do a side project, because they'll trade off the beer and pizza or the basketball on Wednesdays or the time on Saturdays, they will trade it off for doing this entrepreneurial thing.
39:23
So, you know, go find a job, beg your way onto the team, do any job that is necessary in the company.
39:31
Take out the trash, watch and pay attention to the successful, experienced entrepreneur, start side projects, and at some point you'll come up with the one that feels right to you at that point in time.
39:44
Uh, that's, that's my general advice. We've all been in the situation where our flight got canceled. We've all sat in that airport and was like, "Goddammit, like, which other flights could I take?"
39:59
And then you, like, try to call into some, you know, some, uh, you know, customer service agent who's like, "We'll get back to you in two hours," sitting on hold.
40:07
But by the time you get there, you haven't really researched what are all the different options. So, Reroute is that idea. It is for anybody whose flight gets canceled.
40:18
It is the opportunity to understand, what are all my options to get to where I need to get to? And sometimes that's buy a different ticket, sometimes that's ask for the credit from the first place.
40:27
It's understanding that this thing got canceled because of weather, because you can't... Whatever all those c- those permutations are. So that's, that's Reroute.
40:35
Uh, and I think I would probably charge on a per-use basis, right?
40:42
You could have the app for free, but if you, like, you want Reroute to do a bunch of work, it costs 10 bucks or 15 bucks or whatever, you know, for it to do it.
40:50
I think you could probably get access to some of the APIs, or you could do a computer use thing and go scrape a bunch of, you know, booking.com/expedia to go find out all the options.
40:59
So it's a mobile focus on empowering the traveler to reroute when they are canceled. So that's idea one. This is an interesting one because there was a, a... Do, do you ever remember a company called Freebird?
41:13
Yeah, v- vaguely. So Freebird were, they were, they were like a flight cancellation concierge dressed up as an insurance company. Hmm.
41:21
So the idea was two days before your flight, up to two days before your flight, you could pay a fee, and if your flight got canceled, they would then book you on another flight automatically, and you don't have to pay a cent.
41:33
So it was effectively like just, just like insurance, but what killed them was COVID.
41:39
So when COVID hit, there was a lot of disruption in flights, and then all of a sudden they had this massive influx of people that- Hmm...
41:47
needed flights booked, and they couldn't absorb the cost, and they eventually went into, uh, a bad state and were, uh, acqui-hired by Capital One, I think it was. Uh, the, this is great. I think the thing here is timing.
42:03
You know, it's g- like, like insurance, no one wants to buy insurance until, until they need it.
42:07
[laughs] So, you know, the time at which I feel the most pain is when I hear the flight's canceled, and if you can get to me at that time with a, "Here's a solution for you. It's gonna cost this.
42:19
I'm gonna use your Amex points," or, "I'm gonna use whatever points you have to figure out the best way for you to get from A to B. Do you want it? All you need to do is click a button." I think that works really well.
42:29
I think if you don't have that touch point at the point of most pain, uh, I think this is, uh- It feels, it feels like a pro- like a, an extension of a greater tool or a greater umbrella within the airlines or within these big card companies that already have all the integrations
42:46
Yeah, they, they should. And just like many things, many ideas, you're like, "Well, sure Amex should do it," but have they? They haven't. [laughs] You know, and if- should Expedia or Booking or one of those guys do it?
42:59
They should. They might, um, but they haven't, because I watched... I mean, I've traveled a bunch in the last two or three months. I mean, I've watched so many people doing this exact same problem.
43:12
It happened to my nephew last- You know, and the solution, nobody has created this solution yet, or if it's, if it has created, they haven't marketed it very well.
43:23
Well, I was just gonna say, I saw a bunch of people had their flights, uh, canceled coming to New York for the World Cup final. Uh, that's a pretty expensive flight to get canceled.
43:31
Your, your flight might've been 500 bucks, but your ticket was probably 30K [laughs] or, or something. So yeah, that's, uh, that's where things get a bit dicey, I think. Uh, okay. We probably have time for your last one.
43:41
One more. Okay, this one's called Forge, and this is very much in the Half Baked world and very much in the Lev world.
43:50
I have ideas all the time, and there wasn't a particularly good way to capture the ideas, so I built a Claude project called Forge.
43:58
And I literally just pick up my phone, and I open up Forge, and I just tell about the idea, exactly what we're doing here.
44:04
And it does a piece of what Half Baked would do, as you all do with the newsletter, which is what is it, light level of competitive research, and it now has a score on how good is this idea and how buildable is this idea.
44:20
And in some ways, it's very similar to what you all do, but at least it's in my hand all the time, and like my daughter had an idea the other day, and I'm like, "Great, let's just put it in, let's just put it in Forge."
44:31
So now I have my own personal...
44:32
And I b- I put Notion on the back end of this, so now I have a Notion database that Forge is writing to that effectively means any one of these ideas that come out of Forge, I click one button, and I put it in Lev, and I keep moving it forward.
44:46
So Forge is an idea capture machine primarily for software ideas. Is there a, is there a world where you take Forge and you use that effectively as, like, a lead magnet to get people into Lev?
44:58
So this is, like, the top of funnel, like, free tool or whatever, store your startup ideas here or whatever, and then when you're ready to pull the trigger on one, Lev, you know, Lev is the upsell.
45:07
Is that kind of a way you're maybe thinking about that in the future? It could be. I mean, you could go do that and, I mean, Lev is kind of already that today. It's just isn't- Mm.
45:15
Lev is not particularly well optimized for the mobile use case. Um, it's also not, it's not particularly well organized for what I would call the voice interface. I mean, Claude's voice interface is so powerful.
45:29
It's so good, you just, like, randomly just start talking at it, and it, like, somehow figures it all out.
45:33
Now, of s- of course, I could sort of put more of this on the front end of Lev, um, but that also, it sort of presumes that it's not just an idea capture machine, you know, which is what Forge is, and Lev is really a validation and operate- creation and operation machine.
45:50
So I, I would probably keep them separate.
45:52
And, you know, for everybody who's thinking about being a founder out there, if you have in your pocket a m- opportunity, when I have an idea, I can at least get it out of my head and get it into the capture, and then from that, I can start to understand, like, "Hey, I, I keep thinking about that idea.
46:10
Let's do more work on it."
46:11
So you have this sort of capture mechanism, not to mention the fact that it has the components of what we would s- consider to be, you know, entrepreneurship, like who's the ICP and who else is doing it and who's already funded and some of the stuff that you were doing while we're talking, where you effectively have search built into it.
46:28
But you have now a list of, "Gosh, I keep coming back to these kinds of ideas." And that, for a founder, for an entrepreneur-y kind of mindset, is just, that's just part of the, that's just part of going to the gym.
46:42
You know, idea, capture idea, characterize idea, and some su- small subset of those you will go try and prosecute.
46:51
If you ever have any, uh, surplus of ideas that you, you really want to get off your chest, feel free to send them our way. We'll throw them in the newsletter.
46:58
We'll get them out to 100,000 people and try and get them out, built into the world. I think that, uh, that could be a good, a good, uh, forcing function for doing that.
47:06
Well, you guys are doing, you g- a great service of it, and if we had... Imagine now that we could take a million people and get them Forge.
47:13
And of the best ideas that they thought, you know, they're gonna try and do some of them themselves, but they're also going to send some of them to you, and they're gonna send some of them to put in front of another founder who doesn't quite have the right idea.
47:28
Or I think what you'll probably find, if we did Forge plus Half Baked at scale, you'd start to build also connections and community.
47:36
You're like, oh, just like you did with Patrick Collison and my idea of Route Protocol, like, I didn't know he had that. He didn't know I had that.
47:43
But both of us, if we got together and said like, "Shit, why are we not starting that?
47:47
Why are we not finding a young founder that we could mentor, that we could nurture, that might have a, you know, medical degree or knows something more about this than we do?"
47:56
Like, w- how do we catalyze those kind of connections and catalyze those around ideas?
48:02
Part of it's idea capture, part of it is matching, part of it is then saying, you know, picking, picking that up and then putting tools like Lev behind it to figure out, you know, is this idea worth pursuing or not.
48:13
Uh, but that's, that's something I hope we can all work on together in an ongoing way. I think the phrase of whether it's worth pursuing or not is the key towards turning this from a
48:27
useful lead magnet in your pocket idea to a potentially very, very valuable business. Like, it's tapping in a problem that John and I see all the time.
48:38
We have- You know, hundreds of emails come in in a week from people with the number one question saying, "Hey, I have an idea. How do I know it's gonna work or not?"
48:45
And this whole premise of validation as a service is...
48:50
I wouldn't say it's, it's, it's, you know, you know, 100% repeatable across every idea, but there is a met- like, a methodology to figuring out whether, you know, signal versus noise, is this idea something that the world needs or not?
49:05
And, you know, one of the most tried and tested ways of doing that is throwing up fake door landing pages, running some paid ads, understanding how the market reacts to your product in a cold way, in a cold capacity.
49:20
And only a very small percentage of the world actually know how to do that, and it's not that difficult.
49:24
So I'd love to understand if there were, there was ways you could grow legs on this idea and say, "Hey, you know, T- TA, your daughter had an idea, um, it's a fiber supplement that people can take, uh, and it's dressed up in a really cool health bar."
49:41
And then the app said, "I actually think this is a great idea too. Uh, would you like me to do a fake door landing page test for you? How much budget do you want to throw into the meta ad campaign?" Uh, "$500." "Great.
49:53
I'll come back to you in a week with the results. Hey, TA, we got three pre-orders, we got two X, and we got five people emailing us, uh, saying that they want this product. Uh, do with this whatever you wish."
50:05
I think a service like that, whether it's productized, whether it's, you know, Elizabeth Holmes, uh, people behind the sheet doing all the work for you in the back end to begin with, uh, I think that's a problem that taps into a lot of people's desires, particularly around the mental masturbation of, "I have an idea.
50:24
I wonder would it work or not?"
50:26
And that's generally the step that people don't take, and I think if we're looking to create more founders in the world, I think a service like that would do the world, and the people that are listening to this, reading the newsletter or using Lev, I think that would do them a lot of good.
50:39
One tradition we have in the pod, TA, and then we will let you go to actually go and build the company, uh, of, of Lev, is that we like to throw one of the ideas that we have into the oven at the end of the episode.
50:51
We're all about creating more founders. You're also all about creating more founders.
50:54
One of the ideas that we've talked about today, we're gonna take, and we're gonna give somebody the opportunity, uh, and a bit of seed funding to start it.
51:01
So of all the ideas that we spoke about, which one stands out to you as the one you want to see brought into this world? Oh, Protocol for sure. And I, and I would love to do it. There's...
51:11
I'd love to do it, and I'd love to see it exist in the world. And like lots of ideas, I'll be, like, a little bit sad when it's not mine, but I'll be a lot more happy when it exists in the world for somebody else.
51:22
I love that attitude. I wish more people had that attitude. TA, thank you so much for coming on. You've been an absolute blast. Is there anything else you want to leave us with? I just had a, I've had a great time.
51:30
I really appreciate.
51:31
It takes a village of, you know, lots of different people trying to create startups, and you guys have, you know, done an incredible job of doing that, and I, I just appreciate the work that you're doing and, um, appreciate you having me on.
Half Baked
Listen on
Spotify
Spotify
YouTube
YouTube
Recent episodes
Brainstorming $100M Startup Ideas
Sep 15, 2026
