Andrew Warner, Mixergy: Selling to AI Agents, Not Humans

Mixergy founder Andrew Warner on why AI agents are the next customers, how Postiz grew to $70K a month serving them, and why SaaS is maturing, not dying.

Andrew Warner, founder of Mixergy, on The AI Report podcast
April 16, 2026

Watch on YouTube · Listen on Spotify or Apple Podcasts

Andrew Warner, founder of Mixergy and creator of The Next New Thing, says the most exciting opportunity in AI right now is building products for AI agents as customers. He tells host Liam Lawson how Postiz, a social media scheduling tool, climbed from a few thousand dollars a month to about $70,000 after its founder made it the tool agents can use, why so many builders are playing with AI instead of chasing revenue, and why he thinks SaaS is maturing rather than dying.

Key takeaways

  • AI agents are a new breed of customer. Many websites and apps still use CAPTCHA to keep agents away, and few products cater to them. Andrew sees the most exciting growth in software built for agents to use. (27:48)
  • Postiz grew to about $70K a month by serving agents. After its founder turned it into the tool agents can use, monthly revenue went from a few thousand dollars to 20, 26, 50, 62 and then 70 thousand, built by one founder and his wife in Thailand. (28:05)
  • Agents don't churn the way people do. Human users post for a while, trail off and stop paying. An agent keeps publishing like the machine it is, which makes it a great customer, according to Postiz founder Nevo David. (29:06)
  • Shiny object syndrome is the trap. Building with Claude Code and Lovable is so fun that builders, Andrew included, stop asking who else cares about what they make. (13:12)
  • Revenue proves you are helping. Andrew's mission is to help AI builders find the money, because charging connects you to the customer and forces you to prove value. (17:55)
  • SaaS is maturing, not dying. You can vibe code your own tools, but most people won't want to maintain them, just as Andrew used his new bread maker twice. Expect more competitors and harder wins. (46:57)
  • Ask what an agent should be using. Andrew compares this moment to the moves from desktop to web and from web to phone, and suggests repackaging products for agents, or making them agent first. (57:12)

Why build products for AI agents instead of humans?

Andrew Warner says software agents are a new breed of user that few companies cater to, and they are consistent customers. Many websites, apps and tools still use CAPTCHA and similar barriers to keep agents away. Postiz, a social media scheduling tool, did the opposite and set out to become the tool agents can use.

He points to Basecamp too: founder Jason Fried, whom Andrew describes as a minimalist who doesn't love adding features, has said he is adding tools to make Basecamp the project management software agents can use. The appeal is retention. As Postiz founder Nevo David told Andrew, a human user may post on Twitter and Instagram for a while, then trail off, churn and stop paying. An agent has a system that gives it creative input, and it keeps publishing.

Andrew adds that agent spending may be even less visible than a SaaS fee on a statement, because the agent signs up and the payment keeps going out. He had to set up a separate card in Ramp, with a clear limit and a deadline, for the software his agents buy, after usage of a data API suddenly ran wild while Claude Code was building.

How did Postiz grow to $70K a month?

Andrew Warner says Postiz took off by making it easy to post on social media through an agent. Whether someone runs OpenClaw, Claude Cowork or a Paperclip agent, it can connect in, so a request such as "take my YouTube video, cut it up into shorts" and post them to Instagram becomes a simple automation.

Andrew watched the founder share the numbers on Twitter: 20, then 26, then 50, when Andrew expected a plateau, then 62 and 70. The homepage headline calls it "your agentic social media scheduling tool", and when people on X talk about what their OpenClaw agents are doing, they often recommend it, helped by an affiliate program that pays them for referrals. Andrew notes the name is confusing (it is spelled P-O-S-T-I-Z) and it is growing anyway.

What does Zapier's origin story teach founders?

Andrew Warner says the best way to start a business is to find a problem so big that customers want to pay even for a rough solution. He learned it from Wade Foster, who found a post Andrew had written on Stack Overflow about getting PayPal buyers into his CRM. Months later Wade got in touch, and ended up solving a different problem for Andrew: adding a newsletter signup checkbox to his contact form.

After seeing it work on a screen share, Andrew paid Wade before Zapier had a product or a business, and became its first customer. Wade saw it as market validation and went off to build the company. Andrew still has his original Zapier account, and every team he is on has multiple accounts.

Why are AI builders stuck on shiny objects?

Andrew Warner says building with tools like Claude Code and Lovable is so fun that builders, himself included, stop asking who else cares. He finds interviewing people in AI harder than it was in the early days of COVID, when the entrepreneurs he spoke to were thinking about how to be useful to customers.

That is why his mission now is to help AI builders find the revenue. He compares ignoring money to early startup founders who said they only wanted to change the world, and who he knew were headed for pain later. Revenue lets you step away and travel, and it connects you to your customer: people will try anything that is free, but once it is paid, you have to prove value.

What is in Andrew Warner's AI tech stack?

Andrew Warner builds with Claude Code inside Visual Studio Code, with a project's files on the left, their contents in the center and his conversation with Claude on the right. He also writes alongside Claude in a shared doc in VS Code, which he calls magical.

His other daily tools are Atlas, OpenAI's browser, which he prefers to Chrome with Gemini because he can chat about any page and limit answers to a transcript he uploads; Claude desktop, which he added reluctantly to stay minimal; and WhisperFlow for dictation, which he likes for its speed but not for letting the music he is playing bleed into what he dictates.

He keeps a separate laptop just for OpenClaw, and calls it both the most impressive and the most underwhelming thing in AI right now: it can control a browser, code, send daily reports and queue up email, but it needs constant adjustment. Outside work, he set up an OpenClaw agent as a personal shopper that recommends clothes and takes his feedback in a Notion database.

Is SaaS dead?

Andrew Warner says no: SaaS is becoming more mature, with more competition, but it is not dying. He explains with a bread maker. When his family kept running out of sandwich bread, he bought one, then used it twice, because having bread delivered or picked up was easier. In the same way, he can vibe code a lot of apps, but he doesn't want to keep improving and maintaining them, and companies would rather buy software and send their questions to the vendor's help desk.

He expects more custom software built for individuals and teams, and more players overall. That makes wins harder. Buffer's founder Joel could launch with a post on Twitter and a wait list, while Postiz, in essentially the same business, faces far more competition. Andrew thinks many dictation apps will get "slaughtered" unless they really stand out.

Liam's view is that if intelligence becomes a utility you pay for, audience, distribution and sticky workflows become the moats. Andrew agrees that people often keep software because switching is a pain, not because it is the best, though he notes that an agent could now copy your settings into a new app for you.

Which everyday tasks should AI agents take over?

Andrew Warner thinks much of what people do should become agent to agent. The agent built into his browser handled an order of books for two people, working from a screenshot of their addresses and stopping short of hitting send as he asked. It was slow, but it worked, and he wants agents to handle research such as looking up guests on LinkedIn.

His thought experiment: just as companies once moved desktop software to the web and web products to the phone, ask what you use that an agent should be using, and what could be repackaged for agents. Examples from the conversation include a newsletter written for agents, a chat app that lets agents talk to each other securely, and a document tool rebuilt to be agent first, for agents and humans.

Liam describes his own setup: a recurring task in the Claude desktop app checks his calendar each day and, for new podcast bookings, uses Perplexity Computer and Claude to build a research brief in Notion. Andrew plans something similar, with a private calendar entry linking to a Notion research doc for every meeting. He also admits that, net, he and others are not saving time yet: they are exploring the software and playing with it. He thinks that is how you learn, and that agents will eventually take over the equivalent of sweeping the floor, such as tagging and naming YouTube videos.

How did Andrew Warner build and sell an email newsletter business?

Andrew Warner started with email newsletters, betting he could pay about 10 cents per subscriber and earn a penny a day from each. Signups proved harder, so he built online greeting cards with a pre-checked newsletter box, paid site owners to add them, and then added co-registration: selling checkboxes for other newsletters, whose publishers paid a buck and a quarter per subscriber or more.

When the market turned and sponsors went out of business, revenue fell from maybe three million a month to half a million. Burned out, he paid off the company's debt, sold it with his brother, who was his partner, and traveled. Mixergy grew out of in-person events he later hosted in LA: when Microsoft sponsored them, someone there suggested he show what he was doing online, and interviewing the people who came became his focus.

In his words

"Look for a problem. And if the problem is big enough, even if you don't solve it well, the customer is gonna be so happy that they're going to want you to take money." (12:04)
"Charging and making money shows that you're actually helping people." (17:55)
"They happen to be software. And the thing is, nobody's or few people are catering to them." (28:59)
"If I get an agent, the agent doesn't stop." (29:18)
"And SaaS is becoming more mature. It's not dead." (51:54)

Chapters

00:00 Preview
00:46 Introduction and welcome
00:50 Andrew's background in email newsletters
02:24 Selling the business in his mid-20s and traveling the world
06:12 Starting Mixergy and doing 2,500+ founder interviews
09:55 The founders Andrew admires most: Wade Foster and Zapier's origin story
12:30 How solving problems for free changes your career
13:12 AI's shiny object syndrome: building for fun vs. solving real problems
14:59 Andrew's mission: helping AI builders find real revenue
18:07 Andrew's AI tech stack: VS Code, Claude Code, Atlas Browser, WhisperFlow
22:55 The ideal future of work with AI agents
24:44 What's most impressive and most underwhelming about AI right now
25:36 Building a social listening tool with AI
26:59 Which agent tools are saving or making money
27:48 Postiz: growing to $70K a month by selling to AI agents
29:35 The agent-to-agent marketplace future
31:23 Liam's Claude automation for guest research briefs
34:37 Real-time AI workflows with WhisperFlow, Granola and Dispatch
39:22 Why investing time in AI compounds
40:56 Creating marketing assets in five minutes with Claude
42:17 Vibe video editing: the next frontier for content creators
44:42 Thought experiment: what's defensible when intelligence is a utility?
46:57 The bread maker analogy: why SaaS won't actually die
49:18 What makes software defensible: switching costs and stickiness
52:03 Postiz deep dive: the agentic social media scheduling tool
54:47 Agent-first businesses: newsletters, chat apps and tools built for agents
1:00:00 Where to find Andrew and closing thoughts

About the guest

Andrew Warner is the founder of Mixergy, where he has done more than 2,500 founder interviews, and the creator of The Next New Thing, where he publishes his interviews and newsletter and shares what he is building. Before Mixergy, he built and sold an email newsletter business with his brother. Andrew on LinkedIn · The Next New Thing

About the host

Liam Lawson is CEO of The AI Report and host of The AI Report Podcast. Meet the team

Full transcript

Edited for readability. Quotes above are word for word from the recording.

Liam Lawson (00:46): All right, Andrew, welcome to the podcast.

Andrew Warner: Yeah, it's good to be here.

Liam Lawson (00:50): I am so excited to talk about everything: non-technical founders, agents, Claude, OpenClaw, all of the exciting new announcements. But I want to start with a bit of your background. I was doing research into you beforehand, and I looked at your very first LinkedIn experience, Bradford and Reid. Tell me a little bit about that to set the scene for who you are.

Andrew Warner (01:15): I started out like you, with email newsletters. I realized that if I pay some amount of money to get somebody on my email newsletter, and I could make money from them on a regular basis by advertising to them every single day, then at some point I'll break even, and soon after that everything is just profit. At the time I thought it was going to cost me 10 cents per subscriber, because how much could it cost to add an email subscriber? Now you know how laughable that number is.

Then I said, okay, I'll make maybe a penny per day. So if I have them on my list for 10 days, I will break even on the 11th day. That may not be exactly my math, but it was roughly around there. What I discovered was that it was much harder to get people to join the email newsletter than I thought. And advertising does get better and much more profitable if you can go after big enterprise businesses that pay a lot, and if you can get to such bulk that you're able to change somebody's business. They'll pay you way more per subscriber if you can change their overall business with the subscribers you get them.

Liam Lawson (02:24): How did that finish? I want to hear the story of what happened there. You're in your 20s, you're building this business, and at some point it sells. What was that experience like?

Andrew Warner (02:34): What happened was I realized that I couldn't buy enough ads and convert people at 10 cents each. So I played around with a few ideas. I created these online greeting cards, and I realized that if somebody's sending out a greeting card, they're using their real email address. If I have a pre-checked box when they send out their greeting card to a friend, and the box says also sign up for the newsletter, they would sign up for the newsletter, because people will not uncheck a box. And that took off.

So then I started offering anyone with a website money for adding my greeting card system to their site. If someone on their site sent out a greeting card, they would get paid and I would grow my email list. That really took off, to the point where we would make money pretty quickly on people. But we were spending so much money paying affiliates instantly, as soon as somebody signed up, that I needed to find a solution. Then through a friend I discovered co-registration: as long as there's a checkbox for my email newsletter, maybe I should sell a checkbox for someone else's email newsletter.

That's when the money really started flowing, because there were other newsletter publishers who were willing to pay more than 25 cents. They had funding. They would pay me a buck and a quarter per subscriber, maybe even more. And the economics today are still the same. People are willing to pay a couple of bucks per subscriber because they'll make money on an ongoing basis. So I got a couple of bucks per subscriber, and I was paying the affiliates something like 25 cents every time somebody sent out a greeting card. It doesn't take a genius to figure out that if you're paying an affiliate 25 cents and you're making $2 or more, you're making a profit every time a greeting card goes out. That's where the business took off. And of course, people were joining my email newsletter every time they signed up to one of my co-reg partners' newsletters. So our newsletter was growing and we had this money coming in. It was a nice machine.

Liam Lawson (04:27): And it gets to a point where you're in your mid-twenties and you sell this company, right?

Andrew Warner (04:31): Yeah. We got to a point where a couple of our sponsors basically went out of business. They were paying us because if we could get them a million subscribers, they could go public based on those numbers. Suddenly they had a million members. I would call them subscribers, they would call them members. If you have a million members, the market was valuing them based on membership and growth. They would never say that we were the ones driving it when things were good, until the market just went down.

IPOs weren't so easy, and these companies would go out of business. Either they would owe us money, or they would pay their bills but say they couldn't continue with the contract. So suddenly this business that was doing maybe three million a month was down to half a million. I felt like a loser. I felt like a failure. Half a million a month, in retrospect, is obviously very strong. And I started getting burned out.

We had some debt at the company, because I realized interest rates were low and I could borrow to buy things and keep expanding. Finally, when we got to a point where we could pay off the debt, I said, I think I'm done. So I took whatever profits I could off the table, my brother did too, he's my partner, and we sold the company. I moved to LA, lived on Venice Beach, traveled the world, ran with the bulls in Pamplona, did all kinds of stuff. You did too, huh?

Liam Lawson (05:59): Me too. I'm a big traveler as well. I've spent the last four years traveling around the world, so I feel you on that.

Andrew Warner (06:07): There's something about that. I'm glad I did it before I got married. There's the freedom of where do you want to go tomorrow.

Liam Lawson (06:12): Talk to me about that through line. You start Superniche, you build a successful company, you sell out, you travel for a long time. I know there are some projects in the middle, like volunteering at the Dale Carnegie Institute for a while, but eventually you get to Mixergy and you're doing over 2,500 founder and entrepreneur interviews. Why start that in the first place? Why go broad?

Andrew Warner (06:39): At some point I got tired of just hanging out, and I was looking for something meaningful to do. I remember posting ads online saying I'll help anyone with free coaching. The people who take you up on free coaching randomly tend to be losers who are committed to their lives anyway, so that didn't work out. Then I said I would meet people in person and help them out, and I found that people weren't valuing it enough to even show up. But I was kicking around trying to see if I could be helpful somehow.

I'd been in LA for a while by then, and I said, okay, I'm done just riding my bike and going out. I want to make substantive friends here. I reached out to four of my friends and said, look, the five of us each have friends here. Let's each invite five friends to an event, and we'll end up with 30 people coming to an event who are each connected to somebody.

That was really interesting. I would go to an art gallery and offer to have 30 people come in for drinks. I'd bring the drinks and everything. And it was 30 really good people who knew each other. Of course it grew beyond that. At the event I'd ask, does anyone want to host one of these events too? A few hands would go up, and the four people who joined me in hosting the next one would each invite another five people, so the thing started growing. Then Microsoft decided to sponsor, and someone at Microsoft, a woman named Lynn Langit, said, why don't you show off what you're doing online? The thing that made the most sense for me was to interview the people who were coming, and that became so exciting that it became my focus.

Liam Lawson (08:32): When you first started the interviews, were you focused on any particular niche, or were you keeping it broad?

Andrew Warner (08:40): Because my background was online companies, kind of this B2B, B2C world, that's where I focused.

Liam Lawson (08:51): How has that changed over time? For me, the podcast is the perfect vehicle to explore my curiosity, but I have to keep it within the niche of AI and tech because of what we do as a business. It seems like you could go really broad. Talk me through that.

Andrew Warner (09:13): Yeah, that's exactly right. It was super helpful to have a platform where I could talk to people about anything and didn't have to stick with business. I did have an obsession with what became called the startup world: people who are just starting companies, but with the vision of growing really fast and really big. That's largely where I was. But if I hit on a problem like how do I organize my company, I would do a series on organizing your company. When COVID hit, I did a series on how to deal with COVID, and suddenly I was talking to entrepreneurs who, instead of feeling mopey and sad about it, would grow their businesses based on that.

Liam Lawson (09:55): Out of the founders you've interviewed, who do you admire the most, and why?

Andrew Warner (10:02): There isn't a most, but there are people I admire tremendously. One is Wade Foster from Zapier. Let me tell you how I met him. He had listened to my podcast, and he had a job, and he said, you know what, I see this whole entrepreneurial thing, I want to get into it somehow, I see how accessible it is. He started looking around on message boards for a problem somebody had that he could solve. He saw a post I had up on Stack Overflow, where I said I was trying to get the people who bought from me using PayPal into my CRM so I could send them a handwritten card. At the time I was obsessed with these letterpress cards, and I loved sending people handwritten cards.

I couldn't find anyone to respond to that. Literally months later, he contacted me and said, we're working on this, can I show it to you? I said, actually, I think I've moved on, I found a solution I can deal with, but I have another problem of connecting apps. So he got on a call with me, and he understood the problem. People would use my contact form to send me messages about how much they loved what I was doing with the podcast, and I was trying to grow my newsletter. I wished I could have a checkbox on there they could check to say, I also want to subscribe to the newsletter. There was nothing that enabled that. He said, you know what, I think I could do it. He went away for a little bit and came back with this thing he had created.

I think he expected me to try it on my own. I'm glad I didn't. We did a screen share. He saw how tough it was for me to figure this stuff out, but he also saw how excited I was that I could get it. Suddenly I could have subscribers to my newsletter who loved what I was doing so much that they were filling out my contact form to tell me. So I paid him before he even had a product or a business. I said, I just want you to keep building it. He saw it as market validation, and he went off and built the company. I was their first customer. And now they do much more than connecting apps.

To me, when I think about how you start a business, that's it. Look for a problem. And if the problem is big enough, even if you don't solve it well, the customer is gonna be so happy that they're going to want you to take money. He didn't try to charge me. I sent him money because I wanted him to keep going. I knew this was a janky solution, but I needed it so badly that I was willing to pay for it. That's the amazing thing about entrepreneurship: you solve problems like that.

Liam Lawson (12:30): It's a similar narrative for us, for both me and our CTO. We reached out to Martin, our founder, when he had just started the newsletter, and we were solutions oriented. We said, we know you've just started this newsletter, we know we can provide value for free and help solve this problem, and you don't need to pay us. That willingness to go and solve a problem for someone is why I am where I am now. I'm now the CEO of a company because I asked someone on Twitter if they needed help with something. It's a narrative I see over and over again.

Andrew Warner (13:12): You know what, let's bring it back to AI, because one of the problems I have with AI, and this is not me pointing fingers at others, I'm including myself, is that it is so fun to build with Claude Code, with Lovable, with all these tools. It is so fun that all we want to do is build. The problem I'm having is that I'm not stopping and asking who else cares. I'm stopping and asking, is it fun? Let's do more. What else can I add? It's like we all have shiny object syndrome.

As an interviewer trying to interview people in this space, I'm having a harder time than I had in the early days of COVID. People were much more likely to grow their businesses in that period, and in retrospect it makes sense. Their customers were home more, they didn't have other places to spend money, and they were in experimental mode, so they were willing to try new things. The entrepreneurs I interviewed over COVID were thinking about how they could be useful for customers, how they could create something new and helpful.

I find that when I sit down with Claude Code, I think about what would be fun to do. Maybe that's okay for now. I'm giving myself space to do it. But it's so much more meaningful to ask, where is there a problem in the world that I could come in like Superman and save the day? Then you're not just creating another thing, and you're not just trying to make money. You're really changing people's lives. I've been with Zapier ever since they launched. I still have my original Zapier account, and every team I'm on has multiple Zapier accounts, because it's helping that much. Being able to make that kind of impact is tremendous.

Liam Lawson (14:59): Tapping more on impact for a second: we had a brief email conversation before this about your North Star, why you're doing all this. Tell me more about that.

Andrew Warner (15:16): Partially because I am very excited about all these tools. I've never had my life changed so much, so frequently. There'll be a new app or feature release, and suddenly something I couldn't do before is totally possible. Here's a stupid example, but it's great. Look, I'm not wearing jeans today, right?

OpenClaw comes out, and I say, I can't just do work, because work is going to be really frustrating when it doesn't work out. It has to pay off. With all these things, I try to find some things that are personally fun and some things that are work effective. I'm trying not to wear jeans, and I would like to dress better. So in addition to the work agents, I'm going to create an OpenClaw agent that is just going to be my personal shopper. She will go online, look for clothes for me to wear, recommend them, take my feedback and push me a little bit. That's phenomenal, and it's only because OpenClaw was available and you could do whatever you want with it. It works beautifully. I'm kind of nerdy, so I want the recommendations in Notion, in a database I made where I can give feedback in a table. But still, that's a thing that changes my life. I go from wearing jeans all the time to suddenly not.

The same thing happens with work, and that's one of the exciting things. The second thing was that I would talk to people who were building, and there's something in me that always wants to know where the money is. When I asked them where the money was, they said there is no money. It made me think of the early days of startups, when people would say, I'm just here to change the world, and they didn't care about money. I knew they were going to be in pain later on, because you shouldn't hate money. You should love it all. It's like saying, I want to be married, but we don't have to kiss. You're just not asking for your 10 out of 10. You're shying away from it because you think that if you want that, you can't also have meaningful conversations and support for each other.

It's the same thing here with AI. I think the people who are just building are going to miss out if we don't think about where the revenue is. Revenue does a couple of things. It gives you enough money that, if you decide to go travel and run with the bulls in Pamplona like you and I have, you can go do it and disconnect from the world. The other thing it does is fully connect you to your customer. Anyone will try anything. I've been subscribed to people's junk for a long time: if it's free, sure, what the hell. But once it's paid, at some point you've got to prove value. Charging and making money shows that you're actually helping people. That's why I thought, this is my mission. I want to do that.

Liam Lawson (18:07): Talk to me about your tech stack right now. If I look at your YouTube channel, it's all OpenClaw, it's all Claude Code. You recently had that interview with someone from Paperclip. What does your day to day look like? If I sit down at your computer, what would I see?

Andrew Warner (18:24): My computer is so bogged down that I've restarted everything, and this isn't the one I'm using now. On my day-to-day computer, I have Visual Studio Code, so that I can build using Claude Code. I really like that. I also like it because with VS Code I can see Markdown files as I'm working with them in the center column. On the left I see all my files for a project, in the center I see their contents, and on the right I see my conversations with Claude. I can actually build an app, and I can also write and have it see what it's writing with me in a doc we collaborate on. Collaborating with Claude in a doc in VS Code is magical.

I also cannot give up Atlas browser. It is amazing for me. I went back to Chrome yesterday with Gemini, but Gemini is not the same. In Atlas, if you see an email and you don't understand what the guy's saying, you can hit command, side bracket, and ask what he said, and it'll explain it. While I'm writing a Notion doc, I could say, I don't know what the answer is here, give me 10 ways to finish this sentence. That sparks something in me and I finish the sentence. Gemini will do it, but I don't know if it does it as well as Atlas does. Atlas, by the way, is the OpenAI browser where you can chat with ChatGPT about any page.

Sometimes when I'm writing, I need to give it a transcript of a call or a meeting or an interview. I can say, you're going to give me answers to what I'm writing here, but you can only use this source material, the transcript I've uploaded. I can't do that with Gemini. I can't just upload a doc or a screenshot. So I'm absolutely in love with the Atlas browser. It is so much better than people realize. So I have the Atlas browser, I have VS Code, and I have Claude desktop. I reluctantly added it because I'm trying to keep myself minimal.

Andrew Warner (20:36): And then I have WhisperFlow, which I don't love. It's the main thing we use for voice dictation. I like how fast it is, but I don't love that it won't stop the music or the YouTube video I'm playing, so it bleeds into what I'm dictating, which is annoying as hell. How do they not solve that? I've got to check in with them, because they used to do it.

Liam Lawson (21:04): There must be a setting for it, because I love WhisperFlow. I can listen to whatever I'm listening to, some drum and bass, and just use WhisperFlow, and it's fine. My brain can do both at once. But I understand the need to stop it.

Andrew Warner (21:13): Are you listening to music on your earphones when you do that? Or talking? Maybe you're listening to talking.

Liam Lawson (21:20): I'll be listening to techno or something, and then I'll just WhisperFlow to Claude, and they both work at the same time for me. I don't need to pause.

Andrew Warner (21:30): I do wonder if it knows to disconnect from music, but a conversation it doesn't. Anyway, I don't love it for that reason. I also don't love that I can't tell it: I love paragraphs, I hate dense paragraphs, use more paragraph marks. There's no place for me to easily tell it to do that. And I bought so many others. I like Willow, which does it pretty well. I like Monologue from Every, which does it pretty well. I bought one called Voice Ink, and what I like about it is that you pay once, 25 bucks, and you have it for life.

It's all on your computer. It doesn't send any of your voice to the cloud, and it's so customizable. The one thing it doesn't have, and maybe they've updated it, is the ability to hold down FN when you're talking and let go when you want it to transcribe. I have so many ideas, and maybe I should create my own, but there are a million different tools like that. One thing they all could do, and not a single one does, is show on the screen what people have said, every five seconds. But I see a lot of people building it, and I just don't think it's a great business, because anyone can build it and Apple inevitably is going to make it.

Liam Lawson (22:55): You have a lot of ideas, you've tried a lot of tools and you speak to a lot of founders. What does the ideal future of work look like for you?

Andrew Warner (23:07): Here's where you taste it: when you tell an agent to do something and it does it better than you thought could be done. That is the ideal future. You say, I want you to do this thing, and it understands you well enough to get it done and impresses you. Every once in a while we see a little bit of it. I feel like I've now interviewed so many people who built with OpenClaw. You talked about Catherine, who runs Best Self. She is constantly using OpenClaw agents. Then she created Paperclip to help manage the agents, and she's got them in Slack. She's typical of the entrepreneurs I'm talking to and working with.

The thing is, they spend an enormous amount of time customizing, editing and changing their agents. I talked with Eric Siu from Single Grain. He showed me how his agent is helping him write on a regular basis, and you can see on Twitter that more of his conversations and internal work are now being turned into Twitter articles that get a lot of views. Some get a lot, some don't, but he's willing to fling a lot into the world and have some fall flat for the ones that become big hits.

It's still a lot of work to get this working. This is my OpenClaw laptop. It does nothing but OpenClaw. It's shut because I can't deal with it today. It's just so much pampering and editing. So we see the possibility, but at the same time it's not quite there. That's the future, and it's a future we love so much that we're willing to work really hard to get it.

Liam Lawson (24:44): What do you think is most impressive right now, and what is really underwhelming compared to the media narrative?

Andrew Warner (24:53): I think the same thing is both the most impressive and the most underwhelming: OpenClaw. It's most impressive because it can do so much. It can control your browser. It can code for you. It can send you reports on a daily basis. I'm shocked by how many people love having reports sent to them. It can queue up email for you. At the same time, it's underwhelming. You use it and you realize this is so half-baked, so not quite there. If it was a car, it would stop every block to need to be adjusted before it could go. And yes, it would get you driving, but you'd be far better off walking.

Liam Lawson (25:36): You mentioned you're building an application right now. What is that?

Andrew Warner (25:42): I have one I really like. It's not quite done. I spent a few hours on it Friday, and I'll spend some time on it Thursday, because I'm going to friends' where we do show and tell with what we're building, and I want to have something done. It finds hot topics that I should know about and might want to do an interview about. It's crawling different platforms, finding what people are excited about and looking for those little gems I get so excited about that I want to do an interview about. That's been really amazing to see.

Liam Lawson: Social listening, essentially.

Andrew Warner: Yeah, essentially social listening. I'm a little concerned that at the end of all this I'll discover somebody had built this thing and I just didn't know it, because I've never cared about content. Having said that, I'm building it, I'm learning from it and I'm using it. Then when I build the next thing, I have all this experience to bring to it, so I'm willing to give myself that kind of space. But I am a little concerned. I signed up for 10outof10.com once I created my YouTube channel. I didn't know what to do with it, and I finally just canceled it.

Liam Lawson (26:59): I think Reddit and LinkedIn are the most referenced or most quoted sources, Reddit being the leading one. I use it as well, and I find it a little underwhelming, especially because it just uses YouTube specifically. I find it a bit glitchy, and some of the suggestions it gives me aren't exactly on brand or personalized to me. Something I do want to talk about later in this conversation is the idea of the SaaSpocalypse: why pay for this SaaS app that people spent millions developing when I can code it myself in the space of a week? But what I really want to do is make this tangible. You're speaking with all these entrepreneurs and founders using Claude Code and OpenClaw. Out of the applications that have either saved money or made money with these agents, what are they, and how do they work?

[A connection problem interrupted the recording here, and the start of Andrew's answer was lost. He picks up with the story of Postiz, a social media scheduling tool.]

Andrew Warner (27:48): He then started selling to agents. He said, you know what? Everybody has agents, and all these websites and apps and tools are using CAPTCHA and other things to keep agents away. I'm going to turn Postiz into the tool that agents can use. And dude, it's been taking off.

He went from making a few thousand dollars a month to, suddenly, 20, then 26. He's been posting on Twitter how he's doing. The last time I saw it, I saw it at 50, and I said, okay, he's going to hit a plateau now, because he got all the attention and all the people who have agents will have signed up. Nope. He then showed 62, then 70 this morning. It's just climbing and climbing. And he's just a guy with his wife in Thailand building this thing out. That is an absolute dream.

Most recently I saw that the founder of Basecamp, Jason Fried, not a guy who loves to add features to his software, he's very much in the minimalist build world, said that he's now adding tools to make Basecamp the project management that agents can use. So when you ask me where I see exciting growth, it's that there's this new breed of users. They happen to be software. And the thing is, nobody's or few people are catering to them, and they are consistent users.

As the founder of Postiz, Nevo David, told me: if I get a user to sign up for my software, they might start posting on Twitter and Instagram every once in a while, and then they just trail off, they churn and they stop paying. If I get an agent, the agent doesn't stop. It has a system built in that gives it creative input. It will keep publishing, publishing, publishing, like the machine that it is, and it's a great customer. That is exciting.

Liam Lawson (29:35): How does that shape the future? Are we looking at a mostly agent-to-agent marketplace?

Andrew Warner (29:42): I think a lot of what we do right now should be more agent to agent. Look at all the things you and I do that are pains in the butt, that we do reluctantly and that suck up our time, or that we don't do at all. I'll give you one example from my life. I told you how I like to send note cards to people. I then switched to wanting to send people gifts, especially as my friends were starting to have kids. There's a carrier I love that you put your baby in right up here. Especially as a dad, where you don't have the bonding and closeness a mom might have as she's nursing, you put them right there, you wear them like a shirt, and you get to walk around town with the same freedom you always had but the closeness with your baby. I love sending it to people.

Do you know how much of a pain in the butt it is to get their address and put it into Amazon after you buy it? If you look, Amazon's credit card and delivery process is antiquated. It looks just like it did in 1997. But I gave it to the agent that's built into my browser, and it worked. It took a hundred times longer than it would take me. I wanted to send two people books. I said, here are the books, here's their address. I even took a screenshot of their address from their emails. I said, mail it to them and use this credit card that ends in that. Do everything, but don't hit send. It worked beautifully. This is stuff I wouldn't do otherwise, and agent to agent should do it. It should be my computer talking to Amazon's computer. I should just say I want to do it, and it should get done.

That's one thing. Think about how right now, if I'm doing research, I still have to go to LinkedIn and search for a guest I'm interviewing. You're an interviewer, you still have to do that. Why? We still have all these interactions that don't need a human being.

Liam Lawson (31:23): Well, I don't. As soon as someone's booked into my calendar, I have a Claude task that recognizes that, uses Perplexity Computer and Claude to create a research brief, puts it in Notion and pings me. When I was preparing for today, everything was already prepared for me, automatically, based on calendar bookings. So I don't need to anymore.

Andrew Warner (31:51): Interesting. Tell me about that. What is creating the trigger for you in Claude?

Liam Lawson (31:55): It recognizes the calendar event. For podcasting, the event name says The AI Report podcast, and Claude recognizes that. It checks my calendar at the end of every day. If it recognizes that, it goes away and does the research in a format I like, using tools I like. Then it puts it in Notion and pings me when it's ready.

Andrew Warner (32:20): When you say Claude is doing that, are you telling me Claude can actually get triggered by external events? You're talking about Claude desktop?

Liam Lawson (32:28): Yeah, the Claude desktop app. It's just a task. I set a recurring task to check my calendar at the end of every day and take action based on that.

Andrew Warner (32:36): Got it. So it's not that a new calendar event is the trigger. The end of the day is the trigger, and it looks at what's going on tomorrow. That makes a ton of sense.

Liam Lawson (32:46): Or not just tomorrow, but whatever is on the calendar. It cross-references who it has already created briefs for with the calendar. If there's an event that a brief hasn't been created for, it creates the brief.

Andrew Warner (32:56): Why do you like Perplexity Computer?

Liam Lawson (33:01): Good question. I'm still experimenting with both. I don't really use ChatGPT nowadays. I used to a lot. I've been using Claude Cowork a ton, and Perplexity Computer. The difference is that Perplexity Computer is lacking the memory that Claude has, and it can't take actions the same way Claude can. Perplexity's computer use is still to come out. They just announced it last week. So it's not as autonomous as Claude, but I do like Perplexity Computer for research. I feel its data sources are more precise, and you get access to quality data like PitchBook. So if you're talking about startups or VC, it's usually more accurate and up to date. We'll see.

Andrew Warner (33:54): Okay, I'm going to do that. I'm going to go to Cowork and use only Claude to do this. I love hearing your examples. This is a world where use cases are so much fun. I could listen to you do this all day, but let me just dream about what I'm going to do. I have a separate calendar that's just for me internally. For every person I have a meeting with, I want Claude to do research on them, create a Notion doc for me and put it in a calendar entry that mirrors the calendar entry I have with them. That way it's private, just for me, and all I have to do before a meeting is click that link and see everything I need. That is so wonderful. I'm going to do that.

Liam Lawson (34:37): If you just WhisperFlowed that right now, it probably would have been made in the space of an hour.

Andrew Warner (34:42): You know what? I talked with Wade again. When he wants something, he uses Granola, then just copies it out of Granola, sends it to Claude Code and it gets done. I think you're right.

Liam Lawson (34:53): You don't even need to do that. Granola's another example of how I'm using this. Dispatch came out last week, so I thought, let me play around with this. I had just had a call on my laptop, but I was on my way to the gym. We had just signed with a podcasting agency, and they asked for a bunch of files on my computer: slide decks, partnership decks, demographics.

So I opened Dispatch on my phone and said, look at the most recent call I had on Granola, look at exactly what they need, search through my computer, my Slack conversations, my Gmail and my Notion to find the right documents, draft up a Gmail and ping me when that's ready. I look at my Gmail, and it has taken four different PDFs from files on my computer, found roughly based on what it thinks they are from that conversation in Granola, drafted the email and asked for my confirmation to send it. Because I spoke to my phone.

Andrew Warner (35:58): This is so fun. I wish I had Granola going right now so I could capture that and have it spark creativity for me. What I've been doing lately is getting together with friends and saying, let's see what you've been doing, and I get to see things like that. It's so interesting, because it can do a lot, but until you see examples, your brain doesn't come up with use cases. I think we're limited by what we think it could do instead of being open to what else it can do. I've got to get your phone number before we're done, by the way. I want to text you when something comes out to see how you're using it, because I'm sure you're going to fire me up about it.

Liam Lawson (36:37): My thing with all of this is that I just want to make my life easier. I don't want to have to deal with repetitive bullshit. I don't want to create YouTube descriptions or YouTube tags. I want to be able to respond to emails easily. My ideal future of work is that I come in and I speak to something and it happens. I have an idea and it builds it. I talk to people every day, learn what they're building and think, I can create something amazing. I voice dictate what I want built, and it happens. I don't want to deal with any of the scheduling or logging into different apps. I want to make my life easier, and the imperative I have moving forward with everything I do is: can I spend four hours now to save a hundred hours in the future? Everything I'm doing comes from that.

Andrew Warner (37:32): Truthfully, today I find that net-net we are not saving time. Net-net, we are exploring the software and playing with it. I can't tell you how long it takes me to play with something until I get a solution that saves me maybe 10 minutes a day. Having said that, that is the only way you're going to learn. Eventually you even it out and you say, okay, now I'm at a place where I don't have to spend that much time.

The only problem, Liam, is that everyone else is going to be saving time. When they're all moving faster because they're saving time, you have to keep up, and there's no longer this extra space to sit back and relax. But it still becomes more meaningful work. Every entrepreneur remembers the early days when you do the junk work that doesn't matter. My favorite part of the Ray Kroc book about how he built McDonald's was how every little task he did became a whole role for someone else. He used to have to go and sweep the floors. As an entrepreneur, you have to sweep the floors in the beginning. Then eventually you get a team of people who come in and support you, and you get to focus on the things that matter. I think that's where we're headed. These AI agents are going to take away the equivalent of sweeping the floor, like adding tags or naming YouTube videos so other people can find them, so that we can focus on the things that really matter. That's where it becomes exciting.

And no shame in saying, you know what, there is a toy-ish part of this. But hell yeah. I don't want my life to be the drudgery of Homer Simpson sitting at his nuclear plant pressing buttons. If I could live happily playing with my toys, and if that toy is Dispatch on Claude, so much the better.

Liam Lawson (39:22): I think two things. One, yes, everyone's moving at light speed, but everything you do and everything we're learning here compounds. The more you invest at every stage, the bigger that compounds, so I think it's always worth the time investment. And two, I think we're in a bit of a vacuum. You and I are speaking because this is what we do for work, and we speak to people doing amazing stuff. But I speak to friends, to people on the street, to people I meet, and I know maybe a couple of people in the general population who are using AI to the same degree I am. I don't think it's that many people when you think about it. It's a very specific cohort.

Andrew Warner (40:11): I think that's fair. The majority are still using it like a chat buddy you can talk to and get an answer from, like a smarter Google. Give me one other thing you're doing that is so mind-blowingly good you can't wait to share, that you're doing with AI or having AI agents do for you.

Liam Lawson (40:33): Let's see. I do have one example, but it's not that amazing, so let me give you two other examples, and they're all built around the podcast. First, and these aren't fun, but every Friday...

Andrew Warner (40:48): I can give you a fun one too, if you like. Yeah, give me one.

Liam Lawson (40:56): I record a new script for AI News in 5. We record that on a Friday and post it on a Tuesday. All of that script is done for me automatically. I wake up on a Friday, click on my camera, and the script just uploads. It's all done in the exact style I want.

This one was really cool today. We had a team meeting, and I was on a call with some members of our team who are part of the enterprise side of things. They said, we need to create some copy and some creative. I want five variations of this and five variations of that. I don't really know the format, and it's going to take me about a week and a half. I said, hold on a second. I'd already done this for the newsletter, and my Claude, because of the memory I've fed into it over the past month, knows everything about my business and what I like and what works, because it's connected through every possible connector I have.

So I just pressed function, used WhisperFlow and said, I need you to do what you did for the newsletter, but for this. Five minutes later, bang: 10 different styles of copy, 10 different variations, exactly as you'd want it, with instructions for the designer on what images to create and what they should look like, in one document. I said, you don't need to wait two weeks to test creative and launch it. Go and do that now. There it is.

Andrew Warner (42:17): I see that as a view of the future. I want to get you the thing I wanted to tell you about, but I can't get into that app on this computer because I'm locked out. I try to keep the computer I use for interviews pristine. But I'll tell you what I discovered. There's a Y Combinator company that switched from what they were doing to creating vibe video editing. You give it all your video files.

You know how in Claude you have the artifact on the right and the chat on the left, or in Lovable you have the chat on the left and the website on the right? It's just like that. You talk to it and say, what I want to do is tell this story. It will understand what's in the video and put it on a timeline for you. If you want to make an edit, you can edit the timeline, but that's not the way you're supposed to edit. You talk into it and say, I don't like that you cut off my first word, I want a little more explanation here, and it does that. Then you take it out of there and send it to Premiere, where they're imagining people editing. I think that's a mistake. I think they should just say, okay, bang it out, finish the whole thing right there and export. But they're still thinking this is for people who have Premiere and a ton of video files, to make sense of them before they put them into professional video editing.

Liam Lawson (43:34): That's the dream. It's something I've thought about for a long time. I would love to be able to just speak to something to edit videos, because even giving feedback to my video editor is difficult. There's so much back and forth, it doesn't get done instantly and not exactly the way I want it. I've always wondered whether LLMs are smart enough to understand that dialogue and make edits the way I want them.

Andrew Warner (44:06): So this is getting close, and it's on your desktop, so you don't have to send the files to the cloud. But they're still not trying to do the whole thing, which maybe is actually the right thing to do, because if they try to take it all the way, they're not going to get your vision. But they could do a step beyond putting it on the timeline. Anyway, that's one I like. I can give it to you after this interview. I'm trying to find it on my phone here, and I can't. For a while they told me not to even talk about it, so I wasn't putting it anywhere. Do you want to give me one other one?

Liam Lawson (44:42): I don't, not because I'm gatekeeping, but because I want to do a thought experiment with you quickly. A lot of my conversations are with C-suite execs, and we talk about the narrative of the SaaSpocalypse. I've been thinking a lot about what is defensible moving forward. Where is the world moving? I was sitting with Perplexity Computer the other day and said, use PitchBook to see where all the VC and private equity funds have been investing over the past six months, and spell out possible scenarios for what this means for the future. It goes away, does research, and finds that about 95% of these investments are going into the manufacturing of chips, energy, capacity and inference.

One very common narrative from the CEOs of these companies is that in the future, AI and intelligence will become a utility that you pay for. So I'm thinking through what that means. Maybe that isn't what it becomes, but it's a very common narrative right now. If that's the case, what becomes a moat? If you can pay for intelligence, what is actually valuable?

I think two things. I think audience is incredibly valuable moving forward, because of, one, the data it gives you access to, and two, the access to customers and distribution. And I think platforms that are incredibly sticky, that people can't get off and have to use, become defensible: a workflow so precise and so integral to someone's way of life is one of the only things that is defensible moving forward. I don't really see technology itself as defensible. It's just a thought experiment I'm having. I'm curious to hear your thoughts.

Andrew Warner (46:57): Let me take it away from this for a second, and then I'll bring it back and you'll see why. My wife went to San Francisco for a week, and I realized we didn't have any bread for the kids' sandwiches. We constantly have issues with sandwich bread. So I went to Amazon and bought a bread maker, and I thought, now we're going to have fresh bread and it's going to be great. Flour keeps for a long time, and all I have to do is dump the ingredients in, hit the button, and by morning you have fresh bread. Never run out. I used it twice. It's not that much work, but I used it twice. I don't want to bother with the time when I could have it delivered through H-E-B's app, or get it picked up when we're at Whole Foods. I'm a little ashamed of that, because I have the thing, but I don't want to.

I think the same thing is going to happen with all these apps. Yes, I can for sure vibe code a lot of apps, but I don't want to have to keep improving it and coming back and editing it. Someone's going to have to do it. Why should it be me? Wouldn't it make more sense to have one company whose whole job is to do that? Now that company, because everything is rising in quality, is going to have to do a much better job.

I think all these dictation apps are going to get slaughtered unless they really rise up, because there's nothing in them that's special and different from the others, not substantially. But I think we're still going to see people create apps and companies buy them. Companies don't want to waste time trying to get them right, or even, frankly, handling questions about how to use them internally. They want to send those to a help desk at the software company.

I think we're going to have more of it. Some of it is going to be built just for our custom use. You and I are going to do it, and you might even do it for your team. Some companies will have people on staff to create software for their teams, and it'll be really interesting to have these customized pieces of software that would never have been built otherwise. But I think we're going to end up with a software industry that is just more competitive, with more players.

Andrew Warner (49:18): And what makes it defensible, I think, is the same thing. I don't think we use a lot of our software because it's the best. I think it's because we've got it. Look at WhisperFlow. I have WhisperFlow, and I don't love it. But what's the pain in the ass of switching? I'm going to have to try Willow on the desktop, I tried it on the phone, I'm going to have to try Monologue again, which I tried before, and all this.

Liam Lawson (49:45): Do you use snippets?

Andrew Warner (49:48): No, I don't. I created a snippet earlier today because I'm tired of typing in my email address, but they could all do snippets.

Liam Lawson (49:56): Sure, but that's one of the things that makes that platform sticky. That's what I'm thinking about with defensible software: you need to make the pain of switching so high that the platform becomes indispensable. With WhisperFlow, I now have 15 different snippets, and they're beautiful to have, because I just say my calendar and boom, it pops up, or a million other checkout links.

Andrew Warner (50:21): I think you're right that we're going to need that, but I'll tell you something. If you really wanted to switch, take five screenshots, put them into Claude, and Claude will give you a CSV file. Then you say to Claude Code, even if you use Dispatch from your phone, I want you to go into this new app I've got, some other piece of garbage dictation app, and type these into their system one at a time. It'll do it one at a time, no problem.

I do think we're going to end up with a lot of competition. Before, I would talk to entrepreneurs in my interviews who became multimillionaires within a year. Think about Joel from Buffer, the social media posting app. He now has tens of millions of dollars in revenue. His launch was on Twitter, saying, I created this thing that posts on social media, who wants it? People went and signed up, and there was nothing there. He would contact them and say, you're on the wait list, and why do you want to sign up? Then he understood, created the thing, and they bought it. That was possible back then.

Today you need to really rise up in what you create. It's not going to be as easy, and there aren't going to be as many people who are as rich as Joel. The founder of Postiz I told you about earlier is essentially in the same business as Buffer, and eventually Buffer will serve agents the way Postiz does. Having said that, I don't think he's going to be as rich as Joel, because there's so much competition and it's so much easier for someone else to get there. Maybe that's what happens in a mature market. And SaaS is becoming more mature. It's not dead. There's just more of it. There was a time when every business had its heyday. We just don't have that anymore.

Liam Lawson (52:03): If we think about Postiz for a second, explain to me again what it does. It's a social media scheduling tool? So why is that working? There are a million social media scheduling tools. Why is it better than the others? Why are people using it?

Andrew Warner (52:20): He made it really easy to post on social media through your agent. Whether your agent is OpenClaw or Claude Cowork or a Paperclip agent or some other thing, and God knows there are millions of them, it can connect in. And it's so natural for people to say, create this automation for me: take my YouTube video, cut it up into shorts and post them on Instagram or whatever the social media is. It's easy to do, but you need software the agent can interact with. What Postiz does is aim to deal with agents.

Liam Lawson (52:59): How are you spelling this?

Andrew Warner (53:00): Let me find it. Interesting, right? Are you thinking of using it too?

Liam Lawson (53:06): Definitely. It's the bane of my life. I have a social media team that handles everything, but I'm constantly thinking of spin-out businesses that would benefit from having all of the social media posting automated.

Andrew Warner (53:20): Also, listen to how you and I can't even share this company, because it's such a confusing name. Postiz is spelled P-O-S-T-I-Z dot com.

Liam Lawson: Okay. They've got to work on the name.

Andrew Warner (53:34): Yeah, and it's still doing well. If you go look online at what he's doing, you'll see that regardless of how bad the name is, it's going to grow, and it doesn't matter. At the very top: new, generate UGC video with your OpenClaw. He's constantly emphasizing that he's the site for agents. Look at the headline: your agentic social media scheduling tool. Agent is second only to you in that headline. And this is just the homepage.

What's in the culture is even more significant. When someone is on X, I keep calling it Twitter, talking about how they have their OpenClaw agent doing a thing, they're very likely to refer to his tool as the one you should use to do it. Having said that, I should also say that he has an affiliate program, so these people also get paid an affiliate commission for referring him.

Liam Lawson (54:47): I want to think more about that agent-to-agent idea. It seems like such an incredible business model. If you build something that people need for their agents to use consistently, why would they ever stop using it? Why would they ever unsubscribe if the work is done for them automatically in a way an agent understands? What does that come down to: functionality or the data set? Or is there anything else I'm missing?

Andrew Warner (55:21): That's an interesting point. One of the things SaaS had going for it was that you would be blind to it. You or someone on your team would sign up, and you wouldn't even notice the $25 a month recurring fee on your statement. This might be even more blind, because your agent signs up for it, or you sign up for it for your agent, and the payment keeps going out and you don't see that it's even there. I don't know. I can only tell you about my own personal use.

I do have a problem with it. I buy software for my agents a lot, and I struggle so much with the subscriptions that I had to create a separate card in Ramp that has a clear limit and a clear deadline for use. And subscriptions like Postiz aren't even the big issue, because there you have a clean price. It's things like Apify. I wanted to get data from Reddit and other sources, and I needed an API that would get it for me quickly. I went to Apify and got it, and it works great. I gave it to Claude Code, I gave it to OpenClaw, and I forget what other tools I gave it to. Something went nuts, it must have been Claude Code as it was building, and suddenly the API use went nuts. You have the same thing with Claude Code: your API use suddenly goes nuts. These are definite problems.

Liam Lawson (56:43): Yeah, I built an app in Replit to try to create a podcast dashboard, and it ran through 8,000 Perplexity credits in the space of a week. So it's definitely a problem. But I find it so interesting that there's such potential for building something that is purely agent to agent, never really needing to touch a customer.

Andrew Warner (57:12): Yeah. Here's a little thought experiment. There was a period when you had to ask, what am I doing in desktop software that could be made into web-based software? CRM was desktop software, and Salesforce made it into cloud-based software quickly, but so did many other players. Then there was another period where you asked, what is already on the web that nobody has made for phones? A great example was Instagram. There was Flickr for the web, but Instagram became photo sharing for the phone, and there were some unique things that needed to apply to the phone.

Today I think an interesting experiment, Liam, is for entrepreneurs like us to sit and ask, what am I using that an agent should be using? What are people selling that could be repackaged just for agents? I was doing that myself as I was sending out my email newsletter about the interview I did with the founder of Postiz. I thought, why am I writing a newsletter for humans? It'd be interesting if someone created a newsletter for agents. The very next day after I had that idea, Nat Eliason, who has that Felix agent and is a local friend, created a newsletter just for agents. The idea was, have your agent subscribe to this, and I will keep it updated with the best information it needs.

If something like that can be made for agents, think of everything. We now have a Twitter or Reddit for agents in Moltbook. What else can you do? I know someone who's creating a chat app for agents: Shane Mack. He raised money from a16z for a crypto chat app. It's been doing well. It's built into a lot of crypto tools and wallets you already know. Worldcoin from Sam Altman has it, and Coinbase has it built in as the chat app.

He is now creating a chat experience just for agents to communicate with each other. I asked, why does an agent need that? He said, Andrew, imagine you need your agent and my agent to communicate with each other to buy something for us or work on a project for us. How do you know they can talk to each other in a secure way, without you even having to be in there? I started talking about Telegram, and he said Telegram doesn't do it, and gave me a list of all the reasons why. And you realize, of course, I can't just have two agents talk to each other on Telegram. How do you verify that they're there? How do I not have to be in the meeting? There are all these issues I can't even remember. He's building that.

What else can you think of that you use, Liam, that you might need to turn over to an agent or make agent first? Maybe it's not agent only, maybe it's agent first. Look at Dan Shipper from Every. He took Google Docs and made it agent-first software for agents and humans.

Liam Lawson (1:00:00): Something I'm going to have to think about. My mind is just exploding with ideas.

Andrew Warner (1:00:08): All right, that's a good indication of a good interview and a good conversation. Thanks for doing this with me.

Liam Lawson (1:00:16): Andrew, thank you for your time. I appreciate it. If people want to learn more about you, what you're doing and who you are, how do they go about doing that?

Andrew Warner (1:00:22): I'm totally Googleable. But also, if you want to see what we've been doing, just go to thenextnewthing.ai and you'll see my interviews, my newsletter, what I'm up to and what I'm building. All right. Thanks, Liam.

Get the five-minute AI briefing

Join 400,000+ business leaders who start their weekday with The AI Report. Subscribe free

Join the Newsletter
Inchide fereastra