Daniel Hunter on Patience, Taste, and the Age of the Artist by kirupa (https://www.kirupa.com/me/index.htm) | filed under Interviews with Creative People (https://www.kirupa.com/podcast/index.htm) With AI, going from an idea to a working app can take an afternoon. Getting anyone to notice it still takes months, and that gap is where a lot of builders give up and move on to the next shiny thing. I chat with Daniel Hunter, a writer and builder and the creator of Tech Twitter and Zeitgraphs, about what it takes to keep shipping and then stick with what you ship. Daniel first joined me on this podcast back in 2022 (https://www.kirupa.com/podcast/dan_hunter_content_creation.htm), and this time we talk about patience and SEO, shipping into the void, the architect and artist modes he works in, why he thinks artists are better positioned than marketers right now, agent citations, a newsletter written for your agent, my own AI software factory, and the little plaque next to every painting at the Louvre. To learn more about Daniel: 🌐 Tech Twitter: https://www.techtwitter.com/ 🪐 Zeitgraphs: https://zeitgraphs.com/ 💬 X: https://x.com/danielkhunter ✍️ Substack: https://www.danielkhunter.com/ Watch the interview: https://www.youtube.com/watch?v=fTamX_IByaQ Besides ▶️ YouTube (https://www.youtube.com/watch?v=fTamX_IByaQ&list=PL478wQWRhpfa3g6t2maEDJQRFTBPGxTFe), you can listen/watch on 🎧 Spotify (https://open.spotify.com/episode/76EepxMKqtG60rLY191Nsv) and 🍎 Apple Podcasts (https://podcasts.apple.com/us/podcast/patience-taste-and-the-age-of-the/id1650505117?i=1000803003088) as well if that is your jam! About this conversation Daniel last joined Kirupa on the podcast in 2022, and he picks up with Tech Twitter, a database of what matters in the industry built on his own taste and a few hundred people he tracks, and Zeitgraphs, which turns the day’s news into an Obsidian-style knowledge graph where every node looks like a planet. Kirupa frames the conversation around attention: when anyone can go from prompt to output, the hard part is making something useful and getting it noticed. Daniel’s answer starts with SEO. He barely tweets, yet Tech Twitter brings in 300 to 400 visitors and about 20 new sign-ups a day, because he treats it like digital real estate in an up-and-coming neighborhood and has had the patience to stick with it. Kirupa wonders whether five-minute builds set the wrong expectations, and Daniel recognizes the cycle right away: build something in an hour, post it, give up a month later, and repeat. Even a semi-viral spike is wasted if it doesn’t capture emails you can carry into the next project, and Daniel admits he hops between Grok Bot, Devin Bots, and Claude Duet himself. He splits his own work into two modes. In architect mode, well-defined tasks go to cloud agents, and he trusts their pull requests because the linting, AGENTS.md files, tests, and CI are already in place. In artist mode, he spends hours in tight loops with Claude on his own machine, which is how Tech Twitter’s UI avoided looking like slop. Kirupa confesses that he recently forgot how to set currentColor in an SVG, even though he has a book on this stuff, and Daniel’s take is that if you don’t use it, you lose it, and that’s okay. Daniel thinks we’re entering the age of the artist, where the products that stand out are bespoke experiences that are hard to one-shot, like Farza’s launch videos for Clicky, and he doesn’t buy that marketers are the new engineers. A post he reads aloud, “Good software is a set of strong opinions,” is why he doesn’t think SaaS is dead. Kirupa, who used to work at Salesforce, goes back and forth. He sees software going one level lower into data and APIs that agents consume, and he has stopped renewing Adobe Creative Cloud and Figma Pro now that his agents handle the diagrams, transcripts, and clips. Daniel has no pre-ChatGPT baseline for his search traffic, so 70 to 80 organic visitors a day feel like gravy next to the 1,300 to 1,500 agent citations Tech Twitter gets daily. He waters every channel a little, and his newsletter is really for your agent: subscribers get a custom prompt that builds a daily brief, and a Grok Bot is on the way. Supabase and Vercel Workflows keep his stack simple enough to run in about 10 minutes a day, while Kirupa describes his own AI software factory on an $8 a month LAMP server. They close on offers for the site, a joked-about cease and desist from Elon Musk, monetizing without ads, and the mantra of the design agency where Daniel got his start: “The process is the product.” Even at the Louvre, what stayed with him was the story behind a work of art and the little plaque that explains why it matters. What you'll hear about - Patience and SEO turned Tech Twitter into steady traffic and sign-ups, even though Daniel barely tweets. - Shipping into the void wastes attention unless each launch captures something, like emails, that you can carry into the next project. - Daniel’s architect mode hands well-defined work to cloud agents he has learned to trust, while his artist mode keeps taste-driven work local and hands-on. - With an MVP now minutes away, Daniel thinks artists, not marketers, are best positioned to make products stand out. - Kirupa sees software going one level lower into data and APIs, while Daniel argues that good software is still a set of strong opinions. - Agent citations, a newsletter written for your agent, and a low-maintenance stack show how Daniel diversifies his top of funnel. Jump to a topic 0:00 - Tech Twitter and Zeitgraphs Daniel catches up on Tech Twitter, Zeitgraphs, and Devin Bots, and Kirupa frames the conversation around attention and what’s worth building. 4:33 - Attention, SEO, and Digital Real Estate Daniel explains how SEO brings Tech Twitter 300 to 400 visitors a day while he barely tweets, and why he thinks of it as owning digital real estate. 8:22 - Shipping into the Void They map out the cycle of building fast, giving up, and shipping into the void, and Kirupa compares hopping between new tools to sampling at a food court. 14:31 - Architect Mode and Artist Mode Daniel hands well-defined tasks to cloud agents backed by linting, tests, and CI, and saves hours of local iteration with Claude for the UI. 18:47 - Forgetting the Basics, and Why That's Okay Kirupa admits he forgot how to set currentColor in an SVG, and they talk about letting the basics go while still doing some things, like woodworking, just for the joy of it. 23:26 - The Age of the Artist Daniel says we’re entering the age of the artist, points to Farza’s launch videos for Clicky, and predicts more founders partnering with artists. 27:13 - Good Software Is a Set of Strong Opinions A post about software as a set of strong opinions leads Daniel to argue that SaaS isn’t dead, because people don’t want to maintain their own tools. 29:21 - When Software Goes One Level Lower Kirupa describes software becoming data and APIs that agents consume, and the subscriptions he has dropped now that agents handle his diagrams, transcripts, and clips. 34:30 - AI Overviews and Agent Citations With no pre-ChatGPT baseline, Daniel treats search traffic as gravy, counts 1,300 to 1,500 agent citations a day, and waters every channel a little. 42:30 - A Newsletter for Your Agent Daniel shares his llms.txt Easter egg, a custom prompt that turns his newsletter into a daily brief for your agent, and the Grok Bot he’s building. 47:46 - A Stack Simple Enough to Enjoy Supabase, Vercel Workflows, and an agent watching the logs keep Tech Twitter running on about 10 minutes of Daniel’s time a day. 51:29 - An AI Software Factory on an $8 Server Kirupa describes the GitHub Issues, Copilot SDK, and agent personas behind his own AI software factory, all running on an $8 a month LAMP server. 56:27 - Offers, Elon, and Monetizing Without Ads Daniel weighs going fully open source, turns down offers for the site, jokes about a cease and desist from Elon, and explains why he won’t run ads. 1:00:30 - The Process Is the Product Daniel shares the mantra of the design agency where he got his start, and why the how and the why of building something make a better story than a one-shot prompt. 1:05:11 - The Mona Lisa and the Little Plaque A trip to the Louvre shows Daniel why the little plaque next to a work of art, the part that explains why it matters, makes the experience. Transcript The automatically generated captions have been organized by speaker, lightly edited for clarity, and broken into paragraphs for readability. Names and wording may still contain errors. 0:00 - Kirupa Hi, everybody. In this chat, I talk with Daniel Hunter, one of my favorite product builders, all-around cool person when it comes to thinking about AI and all the future implications of how we can best take advantage of it to create really fun things. The names are familiar to you. Daniel was on this podcast a short while ago, where we just were at the beginning of AI transforming how we build things. Fast forward to right now, AI has not only transformed how we build things, it's probably gone a few steps further. So in this fun conversation, we cover a lot of ground: the latest best practices and how AI is changing how we build software, and also the role that artists and some of the more creative individuals that might have been playing more of a secondary role in the past are now clearly in the driver's seat. It's going to be a fun conversation. So grab a chair, grab a drink, and let's get started. All right, Daniel, great to meet you again after our chat almost a few years ago at this point. Time does fly. And so, yeah, what have you been up to? For all the people who haven't seen the first version of this interview, do introduce yourself briefly, and then just, yeah, what are you doing these days? 1:04 - Daniel Hunter Absolutely, man. Thanks for having me back. And yeah, Daniel Hunter, writer and builder. And I am most known for building a fun—I call it a microsite—called Tech Twitter. And yeah, I've been tinkering with that for a while now. It's a pretty unique database based upon my taste of what actually matters in the industry, and a few hundred people that I'm tracking. And I've been able to kind of build other products off of that. One is called Zeitgraphs, as a play on the zeitgeist. And it's a really visual way to kind of just explore the day's news in a knowledge graph, Obsidian kind of style format. You can just hit the play button and it just goes around the galaxy, and each node looks like a planet. It's a really fun kind of experience. And yeah, man, I've just been tinkering like everyone else is. I don't think my story is really that unique in this moment that we're in. As an engineer, really trying to figure out my new workflow, right? I just saw Devin Bots come out yesterday and was playing with it last night, and it's insane. I can truly see how easy it is for people to push out 50 PRs or 100 PRs a day, right? It's really obvious how that can happen. If you have the right kind of systems, guardrails, QA environments, tests in place, it's pretty clear how you could pull that off and not lose quality. And so, yeah, I'd say right now I'm trying to figure out what skills from different stacks I want to kind of pull together into my own flow for going from zero to one with a new product or a new feature. And trying to make a decision on the harness that I want to stick with for two modes of working. One is like architect mode, and the other one is artist mode. And yeah, we can talk more about that, but that's kind of where I'm at. 3:32 - Kirupa No, I totally do want to talk more about that. I think that the reason why I reached out to you again is that you're right: a lot of people are tinkering, but very few people are actually shipping things and actually iterating on and doing cool stuff. You're one of the rare people who's actually taken all your enthusiasm and creative—you know, creative side projects—and taken some of your best ideas and shipped it and made it something that is usable and friendly. And this kind of has a fun theme I think you and I talked about before. It really is like several things. One is, in this world, attention is the hardest thing to actually keep track of. And now with everybody being able to create things, whether by humans or by agents, it's just hard to figure out what the noise is from all of the signal. And then the ability to create things as well: when it's almost infinitely easy now to go from prompt to output, how do you basically do something that's actually useful for someone? Because our time still hasn't evolved beyond 24 hours. No matter how much we try to compress all these things into it, we still have limited attention spans, which seems to still be the biggest bottleneck in everything we're doing. 4:34 - Daniel Hunter So, to your point about attention, let's unpack that a little bit. Trying to get attention on a social media platform like X or LinkedIn, YouTube, right? Sustaining that is difficult, right? And one thing that I love about Tech Twitter is the SEO that I spend a lot of time on has really paid dividends. And so I'm getting now 300 to 400 visitors a day and about 20 new users signing up to access the full kind of experience. And I'm barely tweeting every day. And I think building a microsite that targets a specific niche that has a lot of SEO potential—meaning people are asking questions on a daily basis to their agents, in traditional search, right, and they are able to basically discover you without you doing the work grinding every day—has always been valuable, but it's so much more valuable today. And there's, I think, endless opportunity to build these kind of experiences and to kind of build out a portfolio of digital real estate. And that's kind of my mindset right now: how can I own a portfolio of real estate on the internet, right, that accrues value? And if you think about buying property in the meat space, right, you're going to try and, if you can, you're going to target the up-and-coming neighborhood in a major city that isn't unsafe, but isn't as attractive as the coolest parts of town. And you're going to invest there, and then hopefully you'll be able to kind of ride it out and benefit from being early or taking a little bit of a risk. And I don't see enough people doing that. I see a lot of people trying a lot of different experiments, which is fine, but they throw them away. I've been iterating on a lot of these things for months, and just sticking with it really pays off. SEO is a long game. You can't just vibe code it and snap your fingers and Google's ranking you and surfacing you the next day, right? Takes a long time. And so you have to have patience. And I think a lot of builders don't have patience. And so they're grinding on Twitter, posting, replying 80 times a day, trying to get their content seen and their project seen. But that's not really sustainable. 8:02 - Kirupa Yeah, no, I mean, it's kind of like the old gold rush mentality. You can either go find gold, or you can sell the pickaxes and the blue jeans and things like that. And if your goal is to just tweet 80 times a day, sure, go at it. But if it's something to build a lasting kind of—like I said, a real estate that you can actually keep for the long run and hopefully have it appreciate, then this is the way to do it. 8:25 - Daniel Hunter Exactly, yeah. And I don't know what it is. I don't know why it's hard for people to stick with a few projects and go really deep, you know? 8:36 - Kirupa You know, the thing is, you have a background in computers and programming and engineering, and I do as well. I'm going to speculate here, and I can be way off base here in terms of assumptions. Because we know how hard it is to build something—in a world before AI came about, we know how hard it is to build, ship, and actually see it reach market saturation and so on—our time frame is not measured in milliseconds. It's actually measured in months to a year in terms of validating something. Whereas I feel a lot of people are coming in right now, they just see the potential of: I see a chat box, I'm going to say, “Build me this thing.” Five minutes later, it's built. And to them, they're like, “Okay, now it took me five minutes to build it, and so I should be able to get success in five minutes or 10 minutes or 20 minutes.” Not in my view, which is like, it took five minutes to get here because of all the abstraction, but it really took six months to get here. And so therefore, I'm going to give it six months of time because there's back end, there's front end, there's magic that goes on. Now at this point, I can't explain half the stuff—the code I'm generating, how it works. But I do know that if I had to know about it, it would still not be five minutes. It would be months of actual effort. And I wonder if that plays a role in how it does. 9:44 - Daniel Hunter Oh, dude! Yeah, dude, that's so good! I think we've just identified the cycle, right? You build something in even an hour, right? That's pretty remarkable. You put it on social media, it doesn't quite get the traction that you had hoped, and maybe you stick with it for a little bit, maybe a few weeks, a month, still not really getting the traction that you thought, and you give up on it. But if you keep doing that, fast forward six months and you've gotten nowhere. Maybe you've gone—I wouldn't even say viral, but maybe you've gone semi-viral over the course of that six months with one of your little vibe-coded projects. But it's like a flash in the pan, right? It's just a spike, but you didn't capture any emails, right? You didn't capture any leads for services that you're offering. And so it was just kind of wasted attention. You captured attention, but you weren't able—well, you garnered attention, but you didn't capture it, I guess, right? You generated attention, you generated views, but you actually didn't capture anything from it that you can carry with you to the next project, right? And so what's really cool about Tech Twitter is, like, people are signing up for a newsletter or signing up to use the app, and I'm not mailing them every day. I have no reason to do that. I'm not trying to annoy people. But if I do have something I want to sell down the road, I have this email list that's just sitting there, right? And so, yeah, I think people think, like you said, that it's going to happen sooner than it realistically should for them. And that's just not realistic. And you get into this rhythm of ship into the void, ship into the void, ship into the void. And you've built a lot of cool stuff, but nothing really meaningful has come out of it, other than you've learned some things, which is great. But I don't know, I think the expectations aren't right. 12:22 - Kirupa No, this goes back to something else we talked about last time as well, which is intrinsic versus extrinsic motivation. And so if your goal is to get the most number of clicks and get short-term fame, I look at it as more of an extrinsic motivation. Whereas if you're more intrinsically motivated, you would think about, for example, how do I make this a sustainable thing? Like, when this experiment doesn't quite pan out, do I collect the newsletter or the email addresses, or whatever needs to be that durable artifact that can then take as learnings and then apply to something else, as opposed to just jumping on whatever bandwagon has to be hot at the very moment? And there's no shortage of bandwagon hopping we're seeing from a lot of people these days. 13:02 - Daniel Hunter Right. And I'm guilty of it. I'm pointing the finger at myself. I'm trying Grok Bot one moment, trying Devin Bots the next moment, Claude Duet, right? You're just hopping between these different assistants, you're hopping between harnesses. And I think there is a healthy amount of that. But moderation, right? 13:30 - Kirupa Yeah, I see. I look at that as more like window shopping. You're just sampling. You're at a food court, a lot of great restaurants there, all offering free samples, and you're like, “I don't want to commit to just one just yet. Let me try all of them and then decide which one is going to be really good.” It just so happens by the time you make one rotation to the food court, a new stall has opened up or the old ones will replace another one. So you're just constantly in the cycle of—the new tools, new models coming up is probably faster than their ability to settle in and choose one. But I don't think it's necessarily a bad thing in that sense. So I think, in your point of view, I call it deep research. 14:02 - Daniel Hunter Ah, okay. Yeah, yeah. 14:05 - Kirupa Yeah, we can spin this in a positive way. You're doing research on the right tool. But yeah, now I've been playing with some of these tools as well. It is hard to keep up because on one day they're great, then another thing comes up to make it even better, and you're like, “Wow, this is a new thing.” A new model upgrade happens, and then you're like, “No, no, I'm going to go back to this one,” or something that you never even heard about. Something comes up out of nowhere and you're like, “This is a thing I'm going to be using instead.” And it's quite impressive. 14:31 - Daniel Hunter Yeah, it truly is, man. And so, going back to my artist architect kind of concept, I do think that the way I'm working now, there's a category of work that I can define very easily and succinctly that can be split up into atomic units and I can just give to a cloud agent. It could be Cursor cloud agents, it could be Devin with their new Bots product. And it just spins up these PRs. I look at the screenshots, I look at the video walkthrough that the QA kind of step provides me, and I approve. And that's it. And I'd say the architect mode in my mind really lends itself to this way of working, because your job as the engineer is really designing and architecting the code base and the QA systems, the testing, and all of these things surrounding the product. So that you build enough trust to allow these agents to just kind of run off and solve problems and to kind of add new things over time. And so a lot of the work you're doing is really about the structure and architecture of the system, right? And that's really fun. Because again, I mean, you've seen a lot of builders in this space talk about, “Oh, I've shipped N number of PRs a day. I shipped 3,000 PRs a day the past month,” and it's crazy. And you're just like, “How is this even possible?” It sounds ridiculous, right? You're just like, “There's no way,” right? But then when you start operating in this mode, you realize, no, it's very easy to ship 100 PRs a day. It's not hard at all, once you have your code base set up in a way that allows you to not blindly trust, but intentionally trust the automation. Because you've spent the time in all these different layers, making sure that you have the linting in place, your AGENTS.md files are well thought out and concise, all of your unit tests and functional tests are there, you have your checks and your CI flow. And then for, obviously, UI changes and updates, you're getting the screenshots, you're getting the video walkthroughs that you can actually look at and improve. And then the artist mode is more like, I have Cursor or Claude Code or Codex running on my local machine, and I'm going to spend six hours iterating in tight feedback loops, going back and forth, really getting to a place that I am excited or inspired with. And that's how I built the UI for Tech Twitter. That was all me really going back and forth with Claude and really kind of figuring out all these different touchpoints and really being obsessed about the UX. And that wouldn't have worked. I couldn't have built the UI for that app just by spinning off cloud agents one by one. It wouldn't have worked. 18:29 - Kirupa You've got the typical rounded corners with like a border on one of the corners and— 18:33 - Daniel Hunter It would have looked like slop. It would have looked like complete slop. Yeah. 18:35 - Kirupa Yeah, it looks exactly like slop, exactly. 18:39 - Daniel Hunter But there's a lot of thought and care over way too many pixels that I'm ashamed to even admit. 18:47 - Kirupa I mean, I think that's probably the only skill that we probably have as software engineers at this point. I always talk about the cycle of like, at one end you have the what and the why of what you're building, then there's the how part. And that's where the whole inner and outer dev loop of the building and all that comes in. And then you have the scaling or go-to-market and the validation part. And that's really the only place now. And look at it, it's like there are three stages. The beginning, we're deciding, “What am I building? Is this worth building?” And then the part at the end, “Is what I've built actually good quality? Does it meet my standards for taste and judgment?” Those are the only two real human skills I feel we need at this point, because the middle part of how does it actually turn from idea to output, the models take care of all of that. At this point, I wouldn't even know how to—and I hate to admit it, but I tried writing a small application the other day just for the fun of it in JavaScript, which is a language I've known for a very, very long time. I was making very basic syntax mistakes, things that had never, ever happened before. Right now, I'm actually debugging this SVG thing where currentColor is a value that is used to make it not be a hard-coded color, but inherited from the outside world. I'm like, “Oh great, how do I set currentColor?” I forgot the syntax for putting the custom property for currentColor. I'm like, “This is embarrassing.” I wrote about this stuff. I have a book on this stuff. I forgot all of it because I've been delegating all these details to agents and they do all of this work. And so now I just went through my Claude and I said, “Just make this a drop-down and choose a color, the SVGs change.” It did it. And I didn't even look at the code. I'm like, there's no point in me looking at the code and how it did it, because it's a skill I don't need anymore. 20:32 - Daniel Hunter Yeah, that's incredible. Yeah, I think if you don't use it, you lose it. And I think it's okay. And it sounds crazy to say it's okay to forget the basics of HTML and CSS. But it is kind of the place we're at. And it allows you to consider things that you otherwise probably wouldn't want to spend time on, or you maybe have to hire additional help to do. There are also things that I don't know if I would even have tried, to be honest. There are things that I feel like we get to explore and learn as creators that we should do just for the joy of it. And then we should figure out, okay, how do I communicate this without any expectation of return? And if you really enjoy working on a problem or working on a specific product, you should continue to do that. And I guess my message for the day, in a way, is: don't give up on the thing that you find really interesting or really care about, simply because you're not getting the feedback. 22:17 - Kirupa Yeah, exactly. It's therapeutic for some people to do woodworking. You can always go to a furniture store and buy the finished piece of furniture. But there's something nice about being able to do something on your own, origami. There are a lot of things where there are ways of not having to do it, but if it brings you happiness, by all means, definitely do it. I guess the part where it kind of becomes complicated is if you're trying to basically compete with others, where the goal is, “I want to create high-quality furniture.” At that point, you have to figure out how do I scale, how do I use the best tools and technologies available right now, so I'm able to meet both the quality, the price, and customization, whatever needs to be there. And I think that's where some of this kind of comes in as well. If I'm doing HTML and CSS just for fun, I think it's totally fine for me to spend weeks learning it, getting back into the weeds of all of this and so on. But if I'm working at a company, or if I'm working for a client and they're needing to ship things at a certain time, the expectation now is that everything is being compressed in how quickly it can go from, “Here's the thing we want to do, let's get it out there quickly.” It would be a disservice if I were to now do it the old-fashioned way. 23:25 - Daniel Hunter Yes, agreed, agreed. And yeah, you can wear both hats, right? You can operate in both modes depending upon the situation. And I think we're entering the age of the artist, where the things that we're going to be enamored with are going to be really bespoke, unique experiences that are going to be hard to just one-shot. Not because the models aren't capable, but because you're competing with someone's creativity. And yes, models can be very creative, but if you think about what cuts through now, even when it comes to product launches—one of my favorite entrepreneurs, engineers right now is Farza, and he has his product, Clicky, that's doing pretty well. And his launch videos and product videos are kind of legendary in that they really feel just like small documentary shorts, almost. They feel very organic and unforced and pure. And this feels like a guy just having a good time with his friends, you know? And you can't compete with that. 25:09 - Kirupa No, I'm actually watching—I didn't realize, I've been watching some of his videos in the past. I just didn't connect the name with that. Yeah, you're right. 25:14 - Daniel Hunter Yeah, he's incredible. And that is what will continue to resonate. And artists are better positioned for this moment. And you're going to continue to see this, right? You see a lot of hype on Twitter right now about how marketing is everything. Marketing is king. Distribution is king. Marketers are the new engineers, right? I don't think that's true. I think it's artists. We haven't quite seen this yet, but it's so close. And you're going to see, I think, more startups and more founders actually partnering with artists to help bring their story to life and help communicate why their product matters and why you should give them a chance versus someone else. And you're going to see a lot of partnerships between artists and companies. 26:16 - Kirupa No, I think that makes a lot of sense, because in the past we had to build a product and ship it. Just getting your MVP out the door was a huge Herculean effort, which meant that the part where your product can be differentiated in some form, or have those little artistic details that make people fall in love with your product, those would all be P2s and P3s that would never make the backlog for what we're going to be shipping as part of our MVP. Right now, what happens is that almost everybody with an agent and a decent idea of translating that idea into a mini spec or series of prompts can now create the MVP in a matter of minutes. The part that's going to differentiate them is not—let's take note-taking apps as an example. It's not that I can now build a note-taking app. It's how does that app set it apart from other note-taking apps, where my personality or the unique problem I'm going to solve for is front and center, very neatly designed, and the user experience is top-notch. And marketing, all that kind of plays a big role in it, which is where the differentiation is happening. 27:15 - Daniel Hunter Yeah, so there was this—let me try and find this post. Yeah, so Tarek, who used to work at Anthropic, he tweeted the other day, or he tweeted earlier today, he said, “I don't really think most people will really generate their own software. Good software is a set of strong opinions, and I'd rather have something that's been really thought through by someone else.” And that resonated with me because if you think about artists, artists are the most opinionated people on earth. Artists have very strong opinions, right? And I think if you pair artists with engineers, and you have strong opinions with the software that you're building, then I do think there are opportunities in SaaS. I don't think SaaS is dead. Because people don't want to maintain the software, right? They don't want to have to keep tinkering and updating and figuring out dependency issues and bugs. Yes, you can have loops and automation and things kind of improve your homegrown applications autonomously. For sure. It's getting easier to maintain these things. But there's still something to be said about software being more than just a tool that you use, but it being also a community. Almost like a way of thinking about the world and about how you operate day to day. Especially when you get into knowledge management and productivity software, right? It's not just this app that I can reverse engineer in 30 minutes, right? It's much bigger than that. So, yeah. 29:21 - Kirupa You know, I struggle with this a lot. I work at a software company and I work on software, and I used to work at Salesforce, which was all about SaaS. And so I go back and forth on this. But where I'm currently thinking about, right, is if we take a step back: why does this software exist? Why do these products exist? What is the overall goal you're trying to do? Let's take CRM products, right? I'm running a business. I need to know who my suppliers are. I need to know who my customers are. I need a way to reach them very effectively. It just so happens the way we've all been doing it now for a long time is you have a CRM-like UI that looks very similar to other CRM-like UIs, because everyone kind of evolves together in this direction. But what if we could have a faster path to go from “My goal is to have a better understanding of my customers, my suppliers, my business” and have the output be provided in a way that makes the most sense to me, bypassing all of the intermediate stages that are currently there? In that world, is the software even software in a traditional sense? Or is it just a skill file that your LLM or your coding agent knows about, is able to connect things with some MCP tools and so on, and just give you the output that you want? At which point, I'm no longer even thinking about my CRM. And part of what makes me think that is we saw Salesforce announce their integration with Anthropic, where you can go to Claude Code and now Salesforce data comes directly into it. You bypass the traditional UI. Yesterday at the Ignite conference, they did the same thing with Microsoft's Copilot. They said Salesforce data can now go into Copilot. They're basically making their destination no longer be the CRM, but the data they store, and then the output of it is going to be whatever surface the end user is actually using. I think from that angle, a lot of these software companies and a lot of the products that we tend to rely on historically are all going to go one level lower. There might just be a collection of APIs or tools, whatever unique data they have, that will be provided, but it will be consumed through something else, which might just be our various coding agents. Like, yesterday was—last week, actually, was the first time in many years I didn't renew my subscription to Adobe Creative Cloud. That was one of the things that I've always had a subscription to. I did not renew my subscription to Figma Pro. I'm using the free version of Figma right now, because everything I use those tools for, I was now able to find a workaround where the agents take care of it all for me. I spent a lot of time creating very detailed skill files and training my agents on how to draw a diagram the way I would draw a diagram in explaining a particular concept. And they've gotten really good. I'm like, “Okay, this is great.” This is a multiplier in my output, because my creative input is what's deriving the output we're seeing here. It's not AI slop. But I no longer need to be there actually drawing the rectangle, rounding the corners, and adding the gradients, which I would have done in the past. I enjoyed that, but my bigger goal is I also enjoy just being able to simplify complicated things. And now agents have been able to do that without requiring me to jump through all these tools. The few tools I'm still using are Riverside, which we're using to record this right now, and Camtasia to do some of the video, the stitching and all the edits. But the agents are doing so much more of it as well, because after this recording and our session is over, I'm just going to put both your video and my video part of it in Camtasia and export it as a video file. Then I tell Claude—I have a podcast skill—I just tell it, “Go ahead and do the transcript, make sure to adjust the direction of all these things correctly, and make sure this transcript fills in the gaps.” It's not a verbatim transcript, but it's more like it summarizes it appropriately, so it's still 90% of the actual words you said, but not exactly all of it. This would have been a multi-hour effort, or I would have paid three different services, like Descript and all these other tools that I no longer pay for. I just basically pay for my Claude Code subscription or Claude subscription, and it takes care of all these for me. It even creates the TikTok version, and it says, “Here's the highlights.” I'm like, “Give me three clips of 30 seconds of the best material that I can put onto all these networks.” It does that as well. And that to me is huge, because this is something that, if I was the creator of these tools two years ago, that was the business to be in. It's like, how do you make this all work? How do you do the text to speech accurately? How do you get the captions? I used Adobe Premiere for one thing, and that was adding the transcription, the subtitles in that font that appears below it, because I had no other reason for Premiere other than that. And now, no longer needed. Dreamweaver, same reason. I have an old publishing pipeline. I had Claude Code modernize it completely for me using GitHub. No longer needed for that. And I think the more we see these things, it's going to be a case where what people still continue to subscribe to, let's say a couple of years from now, or even let's say six months from now, is going to be the thing that's going to be at least durable for the next few years. And I do have a slightly pessimistic view of software and operating systems and the products that run on them. I don't think they're going to be the big thing. Getting back to something else you mentioned earlier, you mentioned SEO and getting search results. I find myself increasingly using the AI search overview on Google, so I don't even click on links as much anymore. I just ask a question, and 90% of the time it has it provided for me right there. And how are you navigating that? Getting back to your world of Tech Twitter and your side project where you're using SEO a bit, what does the future for that look like? 34:55 - Daniel Hunter Yeah, that's a good question. So yes, there are plenty of questions where I'm not looking at the sources, whether that's Google's AI search or Perplexity. There are a lot of queries that I have day to day, right, where I'm just looking at the overview answer and then I'm good, I'm done. But there are plenty of times where I'm seeking an answer to a question and I actually want to dive a layer deeper. I'll actually open up a few tabs from a few sources that are provided in that overview. And that's one thing I do appreciate, is that Google and Perplexity still do show you the sources. And what's interesting, I think the surprising thing, is I'm not comparing the traffic that I'm getting from organic search to what the world looked like before ChatGPT, right? Because I haven't been working on Tech Twitter for that long, right? And so I'm only looking at the traffic based upon the current calendar year, really, in 2026. And so it's not like I'm The Atlantic or The New York Times and I was getting X number of visitors a day from Google, and then that shrunk by 80% because of the AI overviews. I have no comparison, right? So I see this as just free traffic. It's all gravy, right? It's icing on the cake, you know? And so the numbers aren't huge, but I think on average right now, it's probably 70, 80 unique visitors from organic search a day. But what is interesting is I'm getting 1,300 to 1,500 citations per day from agents. And what's really interesting about that is if you're a micro SaaS company or a solopreneur, right, and you are trying to increase brand awareness and increase the top-of-funnel traffic you're getting, I still think content is incredibly valuable and meaningful. And focusing on long-form writing and the aggregation of meaningful content, organizing the world's information in a particular niche in a unique way, is really meaningful, because your brand is going to be surfaced in these agents. And just being top of mind, I think, is really meaningful. And so if someone's in Claude or in Grok Bot and they're asking a question and you're cited, right? Even if you are able to get 30 hits a day from that kind of source, but you have a high conversion rate, that's worth investing in. That's still a channel that you should focus on. And I think you can even, with Grok Bot as an example, create a custom bot for a specific function or task. And then you could have, say, Analytics by Kirupa or whatever, right? Or Today's News by Kirupa. You're always top of mind, because your name is sitting in their harness, essentially, right? Your company name is right there. And you're just giving information away for free, but you're a step away from someone buying something from you. And that's kind of how I see all of this stuff, right? You're not going to get a ton of traffic from organic search. You're not going to get a ton of traffic from AEO or agent search or citations. But in totality, the sum total of all these different kind of touch points does really help and is meaningful. 40:14 - Kirupa Is it to diversify your top of funnel? 40:16 - Daniel Hunter Exactly, exactly. That, I think, is the most succinct way to put it, and it's a portfolio. Again, you don't want to over-index on one channel. You don't want to get stressed out or frustrated if one area isn't really doing that well. But you just kind of water each plant every day, and how long does it take to water a plant, right? And so it's just like, “All right, this is what I'm doing.” It's part of my routine. And then before you know it, you have a few sales here, a few sales there, consistent subscriber growth, your email list is growing, right? Your name keeps coming up in conversation with people: “Oh, hey, have you installed Kirupa's new bot? It's really dope.” You just want to be top of mind, right? And I think that kind of awareness and brand recognition, no matter what you're selling or what vertical you're in, is really critical. And I think you're seeing a lot of people who were anti-Instagram or TikTok or what have you strongly considering these other platforms now. Because they're like, “Hey, I don't really want to use Instagram, but if I just spend 10 minutes here, 30 minutes there,” again, just watering a plant, that'll pay off. And it does matter. It is meaningful. 42:02 - Kirupa Yeah, no, I think that's very true. And the interesting thing I've also found is that all these services like Instagram and YouTube make it difficult for an agent to post directly to them. So you still have to visit their website and sign in. There's some hacky tools I found there, but I'm like, I'm not going to trust these hacky tools. They might flag whatever I'm doing as “Don't promote this,” because it came out of some random thing there. Speaking of the agent discovery part, have you built an MCP server for Tech Twitter? 42:32 - Daniel Hunter Yeah, okay. So basically, what I did is I have an Easter egg. The Easter egg is you can access the llms.txt route for your agents to be able to basically run queries against the site. I'm giving that for free. You don't have to sign up for it. It's a public page. And you can go to techtwitter.com/agents, and then you can see all the different kinds of ways your agent can interface with Tech Twitter. If you subscribe, then I'll give you a custom prompt. But I'm not emailing you on a daily basis. This prompt just allows your agent to basically set up a daily routine to get the latest news and create a daily brief. So that's a custom prompt that I just give away if you sign up. And that's the newsletter. The newsletter is really for your agent. It's not something where I'm paying you on a daily basis in your email. Your agent is just creating this daily brief that you can have with your morning coffee every day. And what I'm working on now is an actual Grok Bot that people can install. And then you can just ask questions whenever you want. It's basically like your personal tech journalist. That's kind of how I think about it, like if you had a journalist from The New York Times hanging out with you on a daily basis. And you can just ask questions of the day, but also ask questions for the past year. That's one of the most valuable things about the databases. I've been basically gathering content and organizing it and structuring it for the past 12 months. And so you can pick any day and time in the past year, like, “Hey, what was going on back in February this year? What happened on January 14th?” And that's, I think, going to be really helpful. 45:13 - Kirupa Will this bot be living on your Tech Twitter website, or will it be something I can install or somehow connect? 45:18 - Daniel Hunter You're gonna install it in Grok Bot, yeah. 45:20 - Kirupa Oh, wow, very cool. So I can use it from my Codex or Claude or Gemini and then have it—again, it's like at whatever the name of your bot is going to be—and ask your questions. 45:30 - Daniel Hunter Yeah. Yeah. And so I haven't set up an MCP server yet, but I think I'm gonna focus on this kind of bot approach first. But yeah, the MCP server might be the next move. I'm also trying to be mindful of not putting too much of a load on my database as well. 46:10 - Kirupa Yeah, that's a fair one. 46:11 - Daniel Hunter So it's gotten pretty popular, but I'm not sure I want it to be incredibly popular. There's this middle ground where I kind of like that a lot of people who leverage it are in the industry and are kind of obsessed with keeping up, right, and find it helpful to save time. And the UX is really nice for people to just kind of hit the space bar and go through everything. Every single hit on Tech Twitter is worth your time. You hit the space bar and you actually read the tweet. It's not like you're just scrolling like a maniac for an hour at midnight when you should be asleep. So you could spend 10 minutes on Tech Twitter and get more signal than you would in an hour by just scrolling in the app. And so that's the main value problem. Creating an MCP server and having the floodgates open would require me to spend more money on the server, on the database to support that. So I've kind of been a little hesitant. I would have to think about monetization and all of that, and that's a whole other thing. Yeah. 47:45 - Kirupa What kind of database are you using for this? 47:48 - Daniel Hunter It's all Supabase. 47:52 - Kirupa All Supabase, okay. No, that's actually pretty cool. You know, I'm just wondering, could you vibe code a database on your own that would handle the unique use cases that you're running into with Tech Twitter and avoid paying the extra fees that you might incur if you didn't become wildly popular? 48:11 - Daniel Hunter That is a good question. I'm pretty certain I can move to a more analytics-style database that is cheaper and faster. I think what I value the most with the stack that I have is how easy it is for me to maintain the entire system on the side. Like, I maybe spend, I don't know, like 10 minutes a day? Sometimes like no time at all. There are many days on any given week where I'm just going to the site as a consumer, and I have no thought about anything else other than just enjoying the product as everyone else is. And that is really rare, considering this is a side project. I'm not doing this full time, right? And so why can I do that? It's because of the simplicity of the stack and how well it is integrated into the cloud via the cloud agents and coding agents that I'm using on a daily basis. And so if I want to go use, you know, another database that isn't as well integrated, it's something I really have to learn and spend a lot of time investing in, like, the quirks. I know Postgres really well. So if I know Postgres, Supabase is pretty easy for me to navigate, right? But if I went to something else, it's like, I don't know the—you don't know what you don't know. 50:08 - Kirupa Yes, that's fair. 50:09 - Daniel Hunter Right? You run into issues and you're—then I'm like a solopreneur trying to debug a new database, right? And then I get out of that ability to just enjoy the product and not really spend any time. It is fully automated at this point. All of the fetching that happens—like, I'm monitoring between 300 and 400 profiles. I'm going through that list every 10 minutes, and I have, like, a cursor, like a bookmark, that just moves through that list over the course of a couple hours, and then it resets. And I'm basically fetching the latest tweets for every profile, running it through, like, a taste filter, saving what I think makes sense, and then categorizing it by topic. And all that is happening via Vercel Workflows. And I have observability in place and have an agent that's looking at the logs and seeing if there are any issues, and it self-heals. So it's a really nice system that maintains itself. 51:29 - Kirupa Yeah, no, I totally get where you're coming from on this. But as a side quest, though, I do think it'd be fun for you to try that. The reason why is that, the way we were describing earlier, the ability to have your agents find issues, triage them, fix them, and then only surface the ones that you actually care about for your review—otherwise, it gets done automatically. That's just the whole AI software factory pattern that you're seeing coming up everywhere. And I didn't realize it has a term, but I've been doing this for the past few months now, basically having all this being done. And for me, I've always been not averse to using third-party services. Because I started this in 1998, I've known that every service that existed back then does not exist in the same shape or in whatever form it is. I mean, Yahoo exists as a company today, not in the same shape it did back then. So if I bet on their infrastructure or their solutions, it wouldn't have survived. I mean, GeoCities, which they bought, is no longer there, for example, right? The web host I used, Media Temple, no longer there. They got bought by GoDaddy and they got shut down. And so I've always been very allergic to putting anything on third-party services, because my horizon is decades. It's like, how do I make this work? And so the way I have my AI software factory set up is, GitHub Issues is the only thing that is a third-party service, where all the source of truth of my content is. I have a small script that loads on every page visit from a person, and that logs any, let's say, errors, console errors, promises that get rejected, the usual stuff that goes through. And then every 24 hours, I have a typical job that goes through all of my files and finds any random issues on, like, punctuation. I just have it cycling through all these various things. And I have the agent harness, in this case, the Copilot SDK. They have a Python version of the SDK. I have it on my server itself, so all the AI processing—all the inputs go into that particular area, and then it does all of the orchestration and triaging and all the various activities. And so in GitHub Issues, I have four agent personas that basically reply to every issue. For every issue that gets filed automatically or by me—it's a private repo. I haven't opened it in public yet, because God knows what kind of prompt injections will come up if I did it. Again, that's the downside of doing it on your own. There are some things you probably haven't thought about that a well-known attacker probably knows, and a service has established and has solved it. So it's currently a private repo right now. But for every issue that gets either automatically filed or that I file, I can have it just go through it. There's a classification agent that goes like, “Okay, this is a small documentation bug.” Then there's an actual one that proposes the fix and says, “Here's the PR.” And there's one that basically decides, do I bug me to review it, or let's go ahead and check it in and deploy it automatically as well. That whole end-to-end pipeline runs entirely on—it's an $8 a month GoDaddy server. It's a classic Linux, Apache, MySQL, PHP stack, because that one I know will survive. 54:32 - Daniel Hunter That's amazing! 54:40 - Kirupa You know, if meteors come and dinosaurs come back to the earth again, right? I know the LAMP stack will be one of the few things that will remain. It's like the alligator and the crocodiles, or sharks—whatever happens around us, that thing is going to be here for millennia from now. And so that's why I've been so paranoid about using any other service, even though it would have been easier. I even created my own custom dashboard where I have common issues, number of people who visit certain pages or connect to the forums, all of those things. I created my own custom dashboard as well, and it created a MySQL database to handle all of that. I never just created a database. I don't know what anything is inside of it. It's just every 24 hours, they run these checks to make sure that the [unclear] integration, all these things, work fine. And that has kind of given me a certain level of similar flexibility, in your way. There was a time when one of the integrations broke. I just told my Claude instance, “Something broke here, just go fix it.” It found it, it fixed it. And I said, “Add a regression test to make sure it doesn't happen again.” And that's it. After that, I never had to worry about it again. And so I've been on autopilot mode on all this stuff for the past, let's say, two months now, where I don't look at any of it. I have a small thing that tells me the growth in storage space between the previous day and today, because the only thing that can kill this whole thing is my storage goes out of whatever the limited amount I have for my shared server actually is. And that's one of those tribal knowledge things where you're like, yeah, when you're doing your own logging and you're having your own database, there's going to be a point where the storage is going to go completely out, and something's going to start falling apart. I just had that. What's the only safety check I have there? It's just, flag the growth day-to-day on the storage. Everything else, I'm like, I can go to bed and not have to worry about anything totally breaking. 56:28 - Daniel Hunter Yeah, I mean, I will say that I am tempted, because of how little I have to maintain things now, to pull things into all open source—like, it's a complete open source stack, right? And I might do that because, again, to your point about just changing things up so that I can support more traffic without, you know, exponentially increasing pricing—or costs, I should say. I have thought about that. And so, you know, it's been kind of wild this whole year. I'm starting to get questions about, am I open to offers? And I was like, “Offers for what?” And they're like, “For the site.” And I was like, “Oh, no, I'm not selling.” And so it is this really interesting kind of stage where I can invest more time and thought into it and figure out a path to monetizing elsewhere, which has kind of been my plan all along. Because of, you know, trademark and the data that I'm actually capturing, I've never wanted to go down the route of trying to monetize the site directly. I don't want to get into a war with Elon. I joked with a friend, I was like, “Yeah, I feel like if Tech Twitter blows up, I'm gonna get a cease and desist from Elon Musk.” He was like, “If you get a cease and desist from Elon, you're gonna be able to frame that and leverage that for the rest of your life.” 58:25 - Kirupa Yeah, absolutely. No publicity is—you know, there's something like bad publicity, right? This is like the greatest kind of bad publicity. 58:29 - Daniel Hunter That was it. Right, so he was like, “Don't worry about that.” I was like, “Yeah, that's fair.” But I am trying to be responsible, right? And I don't feel, in good faith, great about just putting up ads on the site and making money just by creating a better experience of Twitter. So that's really what it is. And I've always thought, okay, how do I leverage this asset to monetize something else? 59:08 - Kirupa You provide an archive system, right? You can say, what was hot, let's say, X number of days ago, for example. There is a world where, as a free user, you only get what's hot right now. But if you're a pro user, you pay for your service, and you get access to the archives that go back in time to a certain moment, where you get a really nice snapshot of, okay, three weeks ago, just before this major news thing broke out, here's what the 200 people I'm tracking were talking about. And all of them were completely oblivious except for this one person. And so that's the kind of thing that someone who cares about the timeline and all that would find great value in. 59:43 - Daniel Hunter Yes, and that's kind of how it's set up now. You go to the site and you can see the current moment for free. That's that dashboard. And you can just kind of set that up on a separate monitor and just have it up, and it just updates and refreshes on its own. But if you want to drill down into anything, then you have to sign up. It's free to sign up. But that's the current value capture as the creator: I get access to be able to hit you up about something else down the road in exchange for this service that I'm providing for free. And I feel like that's a fair trade. 1:00:28 - Kirupa No, I think that's fantastic. But, you know, you talked about the artist being the one that's going to be the center of control going forward. What do you think about all these cases where now people can just tell an agent, “Make me this and make it look like that,” and then it just does it? And then you're like, okay, what can I do as an artist at this point? 1:00:50 - Daniel Hunter Yeah, so what's interesting is I got my start in tech at a design agency, and their mantra was, “The process is the product.” And that stayed with me throughout my career, going on 11 years later. And one of the reasons why I think the artist is going to own this moment that we're entering right now is because people are going to care more and more about how and why, right? And if you can tell Claude, “Hey, build this,” and that's the how, that's not a very interesting story, right? If you're at a bar and you're hanging out with a bunch of people, and someone's like, “Oh, what are you working on? What are you doing?” “Oh, yeah, I built this thing.” And you're like, “Well, how'd you do it?” You're like, “Oh, yeah, I just one-shotted it.” Like, that's not an interesting story. 1:02:06 - Kirupa No, no, it's not. 1:02:08 - Daniel Hunter Right? It's like, one-shotted it at the end—that's a lousy story, right? The why could still be interesting, even though you want to one-shot it. The why could be, “Oh, I tried all these different alternative options,” or, “Vercel screwed up and sent me a $2,000 bill, and so I decided to build my own thing, and I one-shotted it, and I saved all this money.” Okay, that's interesting. But the why and the how, I feel like, lends itself well to artistry, and the process really being a part of the end product and why people decide to engage or to buy. And that story, right? That journey that someone went on, how they went on the journey, how they went about doing the thing, and what is the origin story? Why did you do it in the first place? And why did you make certain decisions along the way? And then all of that being a part of the launch or the artifact that is presented at the end—that will resonate, right, with consumers. That will resonate with buyers. And that is how people are going to cut through. And I think what's funny is, engineers, I feel like, struggle with documenting the journey, and also struggle sometimes to recognize that certain moments along the way are really interesting. We take our knowledge for granted. We take our ability and skill set for granted. There are things that some of my engineer friends do that they just gloss over. And I'm just like, “Dude, that's incredible!” And they're like, “What are you talking about?” I'm like, “No, no, this one thing, you know?” And so I think engineers actually struggle to even recognize the moments along the journey that are worth highlighting, right? Whereas artists, they're very conscious and aware and sensitive to those moments, and documenting them and using them as part of their story at the end. And engineers just are—we're just like, “Oh, that thing? You care about that?” It's like, yeah, that's really cool, and that's a cool story, and that's a cool moment, or how you solve this particular problem is really interesting. And we're not very good at that. 1:05:09 - Kirupa No, no, I think this is kind of the case. Like, you ask, how was the Mona Lisa painted? And if you said, “Yeah, I just went to a supermarket and bought it,” whereas you had a story that was actually there behind it—that adds the value to what you're ultimately doing, as opposed to just the output. 1:05:24 - Daniel Hunter Yes. Think about when you go to—like, I saw the Mona Lisa in Paris, right? For the first time early this year. And, you know, you go to the Louvre, it's this enormous, sprawling palace and corridors. And it takes you 40 minutes just to walk to the damn gallery where the Mona Lisa is actually sitting. And then you wait in line to get close, right? And you get your picture, right? And then you kind of move on. And actually, my favorite moment was on the side, where they had another replica of the Mona Lisa, but then the story behind it. And next to every work of art, there's always that little plaque that gives you context on, usually, the why. Like, why is this here? Why does this matter? Right? And this thing about going through an art gallery without seeing—without having that context, without seeing those little kind of callouts, right? It wouldn't be the same experience. There's a reason why those things are there. I don't know. 1:06:41 - Kirupa Well, Daniel, I think we covered a lot of ground as always as part of this. I'll make sure that people have links to Tech Twitter and everything else that you are currently working on, so that they can see all the great work that you've done as an experimenting, but taking it over the finish line and then creating a shipping product as well, which, as we talked about earlier, not too many people do. 1:07:04 - Daniel Hunter Thanks, man. Appreciate it. Yeah, it's a good time. See ya. 1:07:07 - Kirupa All right, man. Talk to you in a bit. Conclusion Browse all Interviews with Creative People: https://www.kirupa.com/podcast/index.htm