Podcast episode with Steve Krouse, cofounder and CEO of Val Town, demonstrating Townie, Val Town's AI assistant that generated a full app from a single prompt in under a minute. Topics include how programming changes the way you think, the rise of the non-technical AI developer, how Townie works under the hood including its system prompt, and Val Town's strategy as AI progress unfolds, plus the value of small, dense engineering teams. # Channel Identity Observations — Video Ep. 41 (Steve Krouse Interview) ## Explicit Self-Description ### Show name and host - The host identifies himself as **Dan Shipper** mid-transcript: "So I'm talking to Dan Shipper, or I guess I'm Dan Shipper talking to founder of Valtown." - The outro names the show: **"AI and I"**: > "You absolutely positively have to smash that like button and subscribe to AI and I." - The host describes the show as "this podcast" and references an episode numbering convention ("Ep. 41" in the title). ### Creator claims about the channel - Dan claims **Steve Krouse influenced the podcast's creation**: > "you are actually one of the reasons... you were like, Dan, you write all the time, but I don't have time for that. I want to just listen to what you have to say." - Dan claims a **writing background** ("Dan, you write all the time") and an existing audience predating the show — Steve says: > "I started out as an audience member of yours in the first place... I stumbled upon Dan's blog when I was a freshman in college." ## Observable Presentation Evidence ### Intended audience - The outro positions the audience as people interested in **ChatGPT and AI**: > "filled with pure unadulterated knowledge bombs about chat GPT... a journey into the future with Dan Shipper as the captain of the spaceship." - Conversation assumes audience familiarity with **developer/AI tooling** (Replit, Cursor, GitHub, Anthropic Artifacts, system prompts, evals), yet Dan explains concepts ("Does it have a client and a server?") suggesting a **semi-technical / informed-but-not-expert audience**. ### Claimed scope of content - **AI-assisted software development demonstrated live**: The episode's stated premise is benchmarking AI progress — recreating an app made a year earlier: > "to see the difference between... last year with Jeffrey Lidon... and where we are now as like a... benchmark of AI progress." - **Founder/startup interviews**: The guest is "the co-founder and CEO of Valtown," and discussion covers strategy, hiring, pricing, and competition. - Dan discloses: > "just full disclosure, I'm an investor, small investor in Valtown." ### Demonstrated format - **Live product demos on-air**: The host and guest build an app during the episode ("we can make an app for this podcast and then never use it again"). - **Long-form conversational interview** with personal history, technical deep-dives, and business strategy. - Host references **other products his company makes** (Spiral — "effectively a prompt builder" — and Sparkle), suggesting cross-promotion of his studio: > "says the man who like has the studio where you have like five different products." ### Tone and style - Casual, friendly, first-name basis with guests ("Steve, welcome to the show"); personal-history framing (college friendship). - Recurring host commentary pattern: "I've talked about this a couple of times before on the show" — indicating continuity of themes across episodes. - Outro is notably hyperbolic/satirical in tone ("Dan, I'm absolutely hopelessly in love with you"), contrasting with the substantive interview body. ## Claims vs. Demonstrated Patterns - **Claim**: The show offers AI "knowledge bombs" and cutting-edge AI content. **Demonstrated**: A live AI coding benchmark plus founder strategy discussion supports the AI-technology scope claim. - **Claim** (implicit): Audience size/significance sufficient to attract notable founder guests. **Demonstrated**: Only the guest's presence and investor disclosure; no audience metrics stated. - No curated metadata was provided; all observations derive from title and transcript. # Delivery Observation: "Do 60-Minute Coding Tasks in 60 Seconds—With AI – Ep. 41 with Steve Krouse" ## Format & Structure - **Interview-style conversation** between host Dan and guest Steve, with a cold-open demo clip that is replayed mid-episode (the podcast-app build appears at the start and again around `[850]`–`[886]`). - Recurring segment pattern: **demo → origin story → strategy discussion → rapid Q&A exchange**. - The episode closes with an **AI-generated outro voice** ("You absolutely positively have to smash that like button…"), a delivery anomaly distinct from the human hosts. ## Tone - **Warm and affectionate**: both speakers emphasize long friendship — "Steve and I have been friends for many years at this point," "It's an honor to have you on the show." - **Enthusiastic, wonder-driven reactions**: "That's actually super cool. I love this."; "man, I haven't had one of those like real AI wow moments in a while." - **Casual, profanity-adjacent informality** from the guest: "a big ass system prompt," "we would just like YOLO changes." ## Vocabulary & Recurring Phrases - Guest leans heavily on **"like" as a filler** throughout: "like, basically lets you code in your browser," "like that's a technique you can use," "like we're in the middle." - Speaker self-repair pattern — **immediately correcting or restarting phrasings**: > "If GitHub Gist were fun and AWS Lambda. No, sorry. If GitHub Gist could run and AWS Lambda were fun." > "hackathon, hackathon, which was a whole nother eye-opening kind of experience." - Signature **punchy aphorisms** from the guest: "throw more LLMs at it," "evals are important for your mental health," "caught fast copycatters." - Shared vocabulary co-created mid-conversation: **"vibe coding" / "yellowing it"** ("The official term, yellowing it. Yeah. I think that's great. I want to make yellowing happen.") — Dan playfully adopts Steve's term. ## Teaching & Explanatory Habits - **Analogy-based explanation** is the guest's core method: - Figma vs. Canva for product positioning ("Are you going kind of a tool for everyone or are you building a tool for professionals?") - Mr. Beast for emerging practitioners ("if you were like talking to Mr. Beast, like 15 years ago… you'd be like, this guy's dumb") - Stock market for perceived model quality ("your product quality being like the stock market is like a very new… a new thing") - **First-person narrative pedagogy**: Steve explains concepts through autobiography — the Logo turtle / derivative story ("I was like, Oh, the derivative is just the way you're pointing"). - Host uses **checking comprehension moves**: "let me make sure I understand like the example you just gave," followed by a restatement. - Guest **hands over the floor explicitly**: "maybe you direct me like what you're curious about, or I could just go piece by piece." ## Communication Patterns - **Overlap and completion of each other's sentences** signals rapport: "Exactly." / "Exactly. So I mean, basically…" - Guest frequently uses **hedged, on-the-one-hand framing**: "It's been fantastic and also a little bit confusing," "on the one hand, I'm really excited about it. On the other hand, I'm getting questions that are so bad." - Host pattern of **validating before questioning**: "That's really interesting," "That makes sense," "That's really cool," typically preceding the next question. - **Full-disclosure habit** from the host: "just full disclosure, I'm an investor, small investor in Valtown." ## Uncertainty Notes - Filler-word frequency and analogy habits are well-supported by repeated examples above; **gesture, pacing, and visual delivery cannot be assessed** from transcript alone. - The closing "in love with you" segment (`[3669]`) appears to be a scripted/AI-generated bit; whether it's a recurring show feature is **not observable from this single transcript**. # Observation Record: "Do 60-Minute Coding Tasks in 60 Seconds—With AI - Ep. 41 with Steve Krouse" ## Speakers - **Dan Shipper** — host; discloses he is a small investor in Val Town; co-founder context (Every), self-taught coder - **Steve Krouse** — co-founder and CEO of Val Town ## Key Positions & Claims ### Val Town & Townie (Steve Krouse) - Val Town described via its homepage poem: **"if GitHub Gist could run and AWS Lambda were fun."** - Street-level description: *"I make a website for programmers to make websites"* — a tool for programmers, browser-based, with integrated deploy/scale. - **Townie** is Val Town's AI agent; current iteration most similar to **Anthropic artifacts**; it automates the copy-paste-between-LLM-and-editor workflow users had been doing manually. - Townie works via: user prompt + a large system prompt sent to **Claude 3.5 Sonnet**. Krouse emphasizes Claude 3.5 Sonnet as the key enabler: *"the most important thing that happened was Claude 3.5 Sonnet."* - System prompt contents: where code lives (Deno JavaScript, not Node), no server-starting concerns, platform bug workarounds, and platform features (e.g., OpenAI integration usage). ### Live Demo: Disposable Software - The demo recreates the app Dan Shipper and Jeffrey Lijon (transcript spelling: "Jeffrey Lidon") built a year ago (~1 hour, with copy-pasting and false starts) — a podcast app with timer, notes, and ChatGPT question generation. - Krouse: the app **finished generating before Shipper finished describing it**. - Shipper calls it a genuine "AI wow moment": *"I haven't had one of those like real AI wow moments in a while. I feel like I'm getting a little jaded."* - Shipper notes the app requires **client-server coordination** (client → server → OpenAI → back), which he calls genuinely complex. - Krouse frames it as **"throw-away software / disposable software"** — software is so cheap to make you can build an app for one podcast and never use it again. ### Seymour Papert & Mathland (Steve Krouse) - Krouse's intellectual origin: an after-school coding program, then discovering **Seymour Papert** (mathematician, student of Piaget) and **Bret Victor**. - Papert's question: why are some kids "bad at math" but no kids are "bad at French"? His conviction: kids who seem bad at math didn't grow up in **"Mathland."** - Papert's **Logo** (predecessor of Scratch) was designed to trick kids into being good at math, not to teach coding jobs. - Krouse's example: drawing shapes with the Logo turtle built intuition that later made **derivatives** intuitive — *"the derivative is just the way you're pointing if you're walking along the curve."* He stresses he *"wasn't born this way"* — the environment made the difference. - Shipper adds historical framing: mathematical **formalism** is only ~19th-century; Newton and Galileo mixed geometry and prose; formalism gains power but loses intuition. ### Shipper's Counterpoint on Math & Sequential Processing - Shipper, despite teaching himself to code: *"I'm still terrible at math."* He describes difficulty with **sequential processing** (steps depending on prior steps — skipping steps, wrong order). - He prefers intuitive/approximate reasoning and values AI because it *"fills in that part of my brain"* — handles step-by-step execution while he supplies intuition. ### Non-Programmers Using Val Town (Steve Krouse) - Surprising outcome: non-programmers arrived sooner than expected — *"that's been a long-term ambition, but... people who really don't know the first thing about coding... make stuff like this in a minute or two and their mind is blown."* - Tension: he gets *"questions that are so bad"* — e.g., a user with a trivial syntax error (removed quote) for whom the experience is unusable. - Mitigation: an **"ask Townie to fix it"** button; downside is Townie regenerates all code from scratch. Krouse: *"the solution to these problems is often just throw more LLMs at it."* - Krouse's segmentation: people **"allergic to code"** are not Val Town's customers — *"The product for those people is... Claude artifacts... or maybe GitHub Spark."* ### Figma vs. Canva / Who to Serve (both) - Krouse frames the strategic choice as **Figma (pro tool) vs. Canva (everyone)**; he is tempted by pent-up non-programmer demand but *"really like[s] serving the smartest customers."* - **Shipper's counter-proposal**: a new type of **AI-native programmer** is emerging; they look unsophisticated now but will be sophisticated in 10 years. Analogy: Mr. Beast looked unsophisticated 15 years ago. Chance to build "YouTube for programming" and instill taste in early adopters. - **Krouse's pushback**: distinguishes AI engineers who put in the work from users who *"hit the fix it button"* repeatedly and are *"allergic to the code"* — he is *"skeptical that those people are... going to become the next Mr. Beast."* He wants the future Mr. Beasts *"who want to put in the work."* - Shipper's rebuttals: (1) AI enables **fractured-attention programming** (his situation with kids), previously only possible by hiring people; (2) Mr. Beast-scale outcomes required "a hundred million people uploading random stuff." - Krouse's conditional concession: it depends on **model progress timescale** — if no Sonnet-3.5-scale jump comes soon, human focus/care is the limiter; if it does, beginners get more leverage. Val Town aims to be **"neutral to underlying LLM changes"** — swap the model name and benefit (as Cursor and WebSim did when 3.5 Sonnet shipped). ### Evals & Stochastic Software (Steve Krouse) - For the first 3–4 months Townie had **no evals**; they "YOLO'd" changes and couldn't tell regressions from anecdotes. - **Social contagion anecdote**: the Twitter panic that "Claude 3.5 Sonnet got dumber" — Krouse believes he contributed to a panic where people collectively convinced each other the model declined; *"evals are important for your mental health."* - Acknowledged gap: much is missed by evals; when Claude 3.5 Sonnet (new) shipped, their evals showed no difference because *"we didn't have any evals that were hard enough."* - Shipper resonates: "yolo-ing it" is the norm for new products; Every has added evals to Spiral and uses opt-in prompt-version upgrades (copy-the-spiral model, preserving user edits). - **Rollback note**: Krouse rolled back from Claude 3.5 Sonnet (new) to the old version because they *"couldn't get the prompting right"* for the newer model. ### Engineering Gulf (Steve Krouse) - Big contrast between serious production engineering (local setup, tests, PRs, 10-minute deploys) and Val Town's model where **"every save is a deploy"** in ~50ms. - Krouse himself cannot modify the production app due to local-config overhead. ### Strategy: Fast Copycats (Steve Krouse) - *"We've set ourselves up to be spectators in this race"* — watch Cursor/Anthropic/Bolt invent UI patterns, then adopt the best ones. Core innovation focus remains the **functions-as-a-service platform**. - Townie itself was mediocre (ChatGPT-clone) until Anthropic artifacts showed the correct UI pattern. - **Pivot considered and rejected**: a full-stack app platform ("SaaSify," Shopify-for-SaaS-apps) targeting non-programmers buying credits. Rejected because doing the whole stack (frontend, backend, monitoring, DBs, migrations, backups) means a **toy version of everything** — "mediocre at all of them" — leading to a feared **local maximum** of unsophisticated customers. Instead: do fewer things sharply, expand later. - **Why backend?** Honest answer: it's the opportunity investor **Dan Levine (a16z/Accel — transcript says "Excel"; speaker attributed as their main investor)** saw, complementing his Vercel investment on the frontend. Also personal: Krouse always hated backend complexity. Competitive gaps: AWS Lambda has terrible UX; Cloudflare Workers is *"a wonderful product"* but not social/collaborative/browser-based like GitHub. ### Positioning vs. Foundation Models (Steve Krouse) - Differentiator: Val Town runs **backend compute**; LLM companies won't want to build functions-as-a-service internally and will partner — *"that's the outsized dream."* - Shipper raised the strategic tension of benefiting from AI companies without being eaten by them (e.g., Claude artifacts overlapping Townie). ### Company State & Team (Steve Krouse, Dan Shipper) - ~2 years old; raised a round Feb/March; team of 4, hiring 3 more starting January; Krouse on the edge of his seat about the team dynamic shift. - Krouse: growth is *"going well, but not like as fast as I want"* — envious of **Bolt.new** ($4M revenue in first four months, as stated) and **Cursor**; skeptical of Replit Agents' hype. - Shipper: Replit worked well for his AI chatbot course (education PMF, per Krouse), but Shipper uses **Cursor** day-to-day; his Replit opinion predates Replit Agent. - Shipper names the "trough of sorrow": doing well-but-not-incredible can be psychologically harder than failing outright. Krouse reports recent uptick with this month leveling off, no paid marketing. - **Team sizing**: Krouse's model remains small dense teams (Instagram, Notion, ~Cursor at 15–20); AI makes it *"more tractable for us mere mortals"* but he still needs seven people, not four. - Current infra work: **virtual file systems / multi-file apps** — not parallelizable; *"more people would honestly make it go slower"* (Mythical Man-Month reference). - Usage-based pricing for Townie is pending and important — these AI features are expensive. ## Meta / Personal Notes - Krouse credits Shipper's blog (found as a college freshman) as their origin of friendship, and for pushing Shipper to start the podcast; conversely Krouse began as Shipper's audience member. - Shipper praises Krouse's strategic clarity and willingness to be publicly honest; notes his own studio is "literally called Every, trying to do everything." - Closing outro (unattributed voice, likely AI-generated): hyperbolic like-and-subscribe pitch ending with a declaration of love for Dan — preserved as expressed, not smoothed. ## Unresolved Tensions (preserved) - **Pro tool vs. everyone**: Krouse leans pro; Shipper argues for embracing unsophisticated AI-native beginners now. Not resolved. - **Effortful users vs. "fix it button" users**: Krouse's skepticism vs. Shipper's fractured-attention defense. Not resolved. - **Model progress assumption**: Krouse plans for neutrality but concedes outcomes depend heavily on whether another Sonnet-scale jump occurs. - **Evals blindness**: both acknowledge only measurable changes get taken seriously — a known gap. # Speaker Attribution & Provenance Record **Video:** Do 60-Minute Coding Tasks in 60 Seconds—With AI — Ep. 41 with Steve Krouse **Video ID:** `01a0f15b-04f5-7498-9463-a3f137ea9b40` ## Speakers & Roles ### Dan Shipper — **Host / Creator** - Podcast host ("Dan, welcome to the show" framing reversed: he greets the guest). - Identifies himself as a **small investor in Valtown** (disclosure stated mid-conversation). - States he started the podcast partly at Steve's urging. - Mentions his own studio/product "Spiral" (prompt builder) and a prior "AI chatbot course." ### Steve Krouse — **Guest** - **Co-founder and CEO of Valtown** (confirmed by both speakers). - Long-time friend of the host; met in college (host at Penn). - Primary narrator of Valtown/Townie product details, strategy, and personal history. ### Unattributed Voice — **Promotional Outro** ⚠️ PROMOTIONAL - The closing segment ("smash that like button and subscribe… Dan, I'm absolutely hopelessly in love with you") is a **promotional/engagement segment** spoken by an unidentified voice, not attributable to either named speaker. ## Referenced Sources & Third-Party Claims (as attributed in transcript) - **Seymour Papert** — mathematician/educational theorist; book *Mindstorms*; Logo language (per Steve). - **Bret Victor** — referenced by Steve as formative reading. - **Jean Piaget** — cited by Steve as Papert's teacher. - **Scratch** — cited by Steve as Logo's successor, made by Papert's student. - **Claude 3.5 Sonnet / Anthropic Artifacts / Haiku** — Steve claims Sonnet 3.5 as key enabler of Townie; claims a Twitter "panic" about Sonnet getting dumber. - **OpenAI/GPT-4** — used in demo app question generation (host's inference, "I assume"). - **bolt.new** — Steve cites claim of **$4M revenue in first four months** (third-party claim, unverified). - **Cursor, Replit, Vercel, Cloudflare Workers, AWS Lambda, GitHub Gist, GitHub Spark, Figma, Canva, Instagram, Notion, Mr. Beast/YouTube** — referenced as comparisons; no specific claims beyond those stated. - **Dan Levine (Excel/a16z-context investor)** — named by Steve as main investor, "spiritual co-founder"; also investor in Vercel (per Steve). - **Jeffrey Lidon** — prior episode guest (live-coded app with host ~1 year earlier). - **Simon Last** — mentioned by host as prior conversation partner. - **Jackson** — named by Steve as Valtown team member handling evals. ## Availability & Uncertainty - **Comment availability:** Not specified in source material. - **Curated metadata:** None supplied beyond the title and transcript. - Demo model identification ("GPT four, I assume") is the **host's inference**, not confirmed. - No timestamps present in transcript; none invented. It's just so cheap to make software that we can make an app for this podcast and then never use it again. So there are three parts. One's a timer. Another one is for notes. And then the last one is like a ChatGPT integration where it'll read the notes and then generate more questions. And I'm laughing because it finished the app before you were done describing what the app was to me. That's actually super cool. I love this. I didn't expect to have non-programmers be able to use Valtown so soon. That's been a long-term ambition. There are people who really don't know the first thing about coding and come on to Valtown and make stuff like this in a minute or two and their mind is blown and they're so excited. And that's amazing. Steve, welcome to the show. Thanks so much for having me, Dan. This is a dream to be on your podcast. So for people who don't know, Steve and I have been friends for many years at this point. We met in college. Steve is the co-founder and CEO of Valtown. And Steve will be able to describe Valtown better than I can. But it's sort of like a social coding site. Steve, how do you describe Valtown? Yeah, it's I think harder than most companies because it can do so many different things. The poem we have in our website, I think holds up remarkably well. If you're a programmer, it's if GitHub Gist were fun and AWS Lambda. No, sorry. If GitHub Gist could run and AWS Lambda were fun. Exactly. So I mean, basically like basically lets you code in your browser and share your code snippets and build and compose code snippets together. And you recently released Townie, which is a very cool AI coding agent. And also just full disclosure, I'm an investor, small investor in Valtown. I think it's awesome. You're doing really, really amazing stuff. And I will also say you are actually one of the reasons, one of the biggest reasons I started this podcast because you were like, you know, Dan, you write all the time, but I don't have time for that. I want to just listen to what you have to say. And you had sort of a vision for what this could be. And so I really appreciate you pushing me to do it because it's been super fun. And also just an honor to have you on the show. Thank you. Yeah. Thanks for doing what I want. I think it's important for your audience to know that I started out as an audience member of yours in the first place. That's how we met. I stumbled upon Dan's blog when I was a freshman in college, and he was a junior at Penn. And I reached out, you know what, I should have found that email because we probably still have it. Like the very cute, you know, like I'm a little freshman, please take pity on me and mentor me. I want to be like you one day. So it's wonderful that I've found myself in the content that I was an audience member of. It is wonderful. Wonderful to have you here. I think, you know, one thing that people should know about you is Valtown is sort of like this, the latest iteration of something that you've been obsessed with for a really long time, which is basically like programming and programming languages and how programming languages change how you think and all that kind of stuff. Do you want to give people like just a little bit of a overview of that part of your brain and where Valtown came from and why it's important to you? That'd be great. Yeah, it is a longer history, I think, than most startups. And there are a lot of startups, I think, that look like Valtown today. But like hopefully we have some like deeper historical perspective because of my background. So I think where I'll start the story is what I went to this kind of life-changing after-school program that taught me how to code when I was in middle school. And I like through that program, like fell in love with mathematics and fell in love with programming. And overall, I just like felt like a smarter human. And I knew it had something to do with this coding program. And then in college, I went to a hackathon, hackathon, which was a whole nother eye-opening kind of experience. But one of the things that happened at that hackathon was that I was pointed at the work of Brad Victor and Seymour Papert. And I read Seymour Papert. And it like turns out that it was all on purpose. Who is that? I've actually, I know Brad Victor, but I don't know Seymour Papert. Yeah, yeah. So Seymour Papert was a mathematician and a like educational theorist. So he studied, he was like a mathematician who then decided to study with Jean Piaget, like one of the fathers of like the stages of development kind of educational theories. And Seymour Papert had this question, which is like, why are some kids bad at math, but no kids are bad at French? Like if you're, bad at French is because you didn't grow up in France. Like we know that if you grew up in France, you'd speak French just fine. Like some kids are just genetically bad at math. That doesn't, that didn't make sense to him. He, his conviction was that those kids who seem like they're bad at math genetically just didn't grow up in math land, grow up in a math land. And kids who are good at math had some sort of math land experience growing up. And his experience was like that. He got obsessed with gears growing up and other mathematical objects that were just in his house. Got it. Okay, cool. So that, that makes a lot of sense. So continue, so, so continue with, yeah, continue with the story. Yeah. So he wanted to see, can we make kids good at math? Can we make a virtual math land on this new computer thing to make kids good at math? And what he came up with was actually a programming language, but it was never to teach kids to code. There were no jobs for coding back in the sixties and seventies. It was like, can we trick kids into being good at math through programming? And it totally worked for me. And so ever since I lived it and then read about how it's all on purpose, it totally blew my mind. And I became obsessed with the power of programming languages, programming environments, developer tools for like, for not only the power that they give human beings to like create stuff, but the way that they change the way you think and like actually make you a smarter human in other aspects of your life. This is, this has been the thread. So like that moment in, when I was like 18 years old has followed me for the last 10 years. How do they change how you think? Great question. So yeah, the book Mindstorms kind of goes through this. One of my favorite examples is in, in logo geometry, the way that the logo turtle works. So, so the programming language Seymour Papert made is called Logo. It's the predecessor to Scratch. So if you've ever used Scratch, it was made by his student and there are a lot of similarities. And in both programming languages, the model is that you have like a character on the screen and you can give instructions to move forward or turn and move and turn and move. Uh, and then in particular with the logo turtle, you would have a pen on your feet. So you could say, put the pen down and then give it instructions on how to move and then pick the pen up and then there'd be a shape on the screen. So an example of how like I learned mathematical concepts, I grew up in math land through this is that like, how do you draw a square? How do you draw a hexagon? How do you draw a circle? Like you, and you, you, you build up as you kind of intuit how to do these things. You build up an intuition for, um, what later in life you'll learn as like a derivative in mathematics. So like, I don't know, five years after I like was, was drawing circles with my feet, you know, mentally you learn to anthropomorphize the problem. So like, that's a technique you can use in a lot of problems, like putting yourself into the code, like imagine yourself. Okay. So if I was like this, like programmers do that all the time. So you anthropomorphize the code in this specific example, like it's very geometric. So then a couple of years later, when I was learning derivatives, I was like, Oh, the derivative is just the way you're pointing. If you're like walking along the curve, that's very easy. Like I've walked on curves for a long time. And I remember in the class, everyone, like I was like tutoring kids after that class and everyone was like, Oh my God, Steve's a genius. He got it so fast. I was like helping the teacher on the board, but I knew it. Like I wasn't born this way. Like I was just, I had these like very powerful experiences. That nobody else in the class had. That's really interesting. So like, let me make sure I understand like the example you just gave. So basically if you're programming in this environment and you're, you're trying to make a circle, the way that you end up having to do that is by breaking the circle into like small little segments that are like the angle of the segment changes just a little bit each time so that you end up making a circle. Cause like circles aren't like actually round, like they're made of infinitely small segments. And that is, that's what then becomes a derivative. And so the, the like concept of derivative was like intuitive to you because of that. Yes. Yes. It was so easy for me to put myself on the paper, like my feet are on the paper and I'm walking along the curves. That makes sense. I mean, I think it's such a interesting, um, this is a whole rabbit hole we can go down, but like one of the really interesting things about the history of math is like, it's only in the, I think like the 19th century, um, that we start to get, um, formalism in math where you're using like all the equations and stuff that get people intimidated. Like if you read like Newton or Galileo, like they're kind of, they're interspersing like geometry with, uh, like actual written text. Like there's a lot more of a, it's like where the sort of Renaissance man idea is, is like comes from like, there's a lot more of an integration between all the different, um, senses and ways of thinking to help you understand math. Um, and it was only like later on that we sort of like formalized it, which is helpful because it's like the formalism is really powerful, but you end up losing a lot of the intuition unless you're like a super math genius or you're lucky enough to, to use, um, a programming environment like this. And I think that's what makes people, um, makes it much harder and less appealing for people, which is interesting. I will also say though, like I, as someone who like, I taught myself how to code, I'm still terrible at math. I think there's something it's possible. I haven't learned right, but like, there's something for me about, and this shows up in other areas of my life too. There's something for me about like just sequential processing, processing, like sequences where each step of the sequence depends on the previous step. I'm very likely to a, it costs me a lot to do that. And B I'm very likely to, um, skip a step or get them in the wrong order or like do all that kind of stuff. So any math that I can do that's like more intuitive where I'm like almost sizing things, um, in my head, as opposed to like following a sequence of steps, um, that I can do, but the sequence thing like really messes me up. And, um, that's why I kind of like the AI stuff. Cause like, I kind of get things intuitively. So I kind of know what I want to do and the AI stuff just like figures out how to write the code or like, you know, do things step by step. So I don't have to, I like fills in that part of my brain. Totally. Yeah. Yeah. It makes a lot of sense. Yeah. So you're thinking a lot about the future programming. That's, that's kind of what you're building at Valtown. Um, and I think like a really interesting, um, intro to this episode is about a year ago when this podcast was first starting, um, I had, uh, this guy, Jeffrey Lidon, who I know is a friend of yours and also someone that you really look up to and respect. And the, uh, the, the cool thing about that episode with him is like, we live coded an app together. Um, and we using, uh, using chat GPT and repl it. And, um, at the time that was like pretty new and like pretty cool. And it was like kind of, to me, it was like wild that you could make an app on a podcast, like just like while talking, like that was previously like totally impossible, but it took us like about an hour ish to, to actually do it. And there was like false starts and whatever. So like we got it done, but like, it was not super easy. Um, and I think it would be fun to see how far we've progressed. That was your prompt for me in this episode is to, is to see, um, the difference between, uh, last, last year with Jeffrey Lid about a year ago and where we are now as like a, as a, um, benchmark of AI progress. And I think you're going to use, you're going to use Townie to, uh, which is Valtown's chat bot product to, to do this. Uh, do you want to introduce it for people? Yeah, sounds great. Um, so yeah, Valtown is a website to write code and share code and deploy code very importantly. Uh, and one way to think about Townie is like, it's taking what Dan and Jeffrey did a year ago and just automating the steps. Like we were seeing people doing this with Valtown for like the last year, they would go to like an LLM and ask for some code and paste it and then get some error and paste it back. And it was a crazy thing. And it seems like over the last year, we've all agreed that that's crazy and that they should be deeply integrated into a single product. And that's what Townie is. Uh, we, we've worked through a number of different iterations of how Townie works. The current iteration I think is most similar to anthropic artifacts. If you, if your audience is familiar with that. Definitely. I think maybe let's start with just like the Valtown homepage so people can get a sense for like what the product is. Uh, I read everyone the poem, but I think, um, I could start with like the description I give people when I meet them like on the street and they say, what do you do? I say that I make a website for programmers to make websites. I think that like kind of anchors it really well, that it's a, it's a tool really designed for programmers and it's designed for you to make web things, uh, particularly server side where you can do front end and full stack stuff. And I think, uh, one of the most important distinctions is that it's in the browser, you write the code, and then it also scales with you to deploy and scale. So you don't have to then redeploy it somewhere else. So it's like all integrated and drastically simplifies the programming experience. Yeah. Cool. Okay. Great. So this is Townie where we can describe in English what we want to build. Uh, I'm going to paste in and start running the app that you and Jeffrey made. And like, as it's coding, I'll describe what's going on. Uh, so the, we're making an app for this very podcast we're on it. The idea from your episode with Jeffrey was that it's kind of throw, throw away software, disposable software. It's just so cheap to make software that we can make an app for this podcast and then never use it again. So there are three parts. One's a timer. Another one is for notes. And then the last one is a, like a chat GPT integration where it'll read the notes and then generate more questions. And I'm laughing because it finished the app before you were done describing what the app was to me. That's actually super cool. I love this. Um, man, I haven't had one of those like real AI wow moments in a while. I feel like I'm getting, getting a little jaded, but that was definitely a wow. Um, uh, because you know, yeah, we made this app with Jeffrey a year ago. It took us about an hour and this, and there was a lot of copy pasting back and forth or, you know, whatever. And you just like made it with just like pasting in a description of, of what we did in the episode. And it's just like here and it works. Wait, can you, can you make sure all the stuff works? Like what's the generate, does it generate questions button work? Yeah. Yeah. So I'm talking to Dan Shipper, or I guess I'm Dan Shipper talking to founder of Valtown fast platform for, I don't know, whatever, whatever, that's fine. Uh, generate questions. So you typed into the interview notes. We have this whole podcast set up. You typed into the interview notes. I'm Dan Shipper talking to Steve Krauss, founder of Valtown. Um, and then you it generated, it used, um, uh, GPT four, I assume some, some AI model to generate questions. And it says like, you know, here are three questions to ask as a founder of Valtown, what was the original vision? So like, it's actually there, you know, reasonable, reasonable questions. It's not totally, not like totally off. Um, and what's really interesting about this is like, so, you know, obviously we just saw the whole thing just like was made in with, with a single prompt. But, um, what's interesting is this requires, um, a client and a server to be working together. Um, because like in order to, I assume in order to have the generate questions button work, it's like sending it from the client to the server, the server sending it to OpenAI and then OpenAI sending it back. That's hard. Like that's, there's some, there's some complexity here. Um, so I don't know. It's, it's cool. Thank you. Thank you. Um, yeah, I can kind of explain any part of it. Uh, maybe you direct me like what you're curious about, or I could just go piece by piece. Well, I guess like, um, tell me at a high level, like what is, um, how does this work and how is it able to fill in all the gaps there? So there's no copy and pasting. And, and, and am I right? Does it have a client in a server? Like how does it all work? Okay, great. So from a high level, the way it works is, um, we take your prompt and then a big ass system prompt that explains how Val 10 works. And we send that off to cloud three, five sonnet. And I think it's very important to underscore how much of an enabler cloud three, five sonnet is when you and Jeff were doing this with, I don't know, chat to be three or 3.5 or four, or even the current chat to be models, we wouldn't get something this good probably, or it would take a lot longer. Yeah. Um, yeah. Cloud three, five sonnet is like, like what's changed since when you and Jeff did it, like a big part of it is we're, we're, it's all integrated in one tool. Uh, we spent a lot of time getting a good system prompt. And then the most important thing that happened was cloud three, five sonnet. Right. Right. That makes sense. Oh yeah. Let's look at the system prompt. So like, what are the core components of the system prompt that like, that made it start to work? Yeah. Um, there are a lot of iterations and if you want, um, if anyone wants a really deep dive, here's like how we built the original prototype and it's changed a lot since then. And then we have a, we just published a new article about like how, how we're running it in production. So we have a lot of notes for people who want to build their own, but, um, at a high level, how this system prompt works is it explains where the, where the code that we're asking it to write is going to live. We tell it that it's running in Dino JavaScript, not no JS, not client side JavaScript. We tell it that it doesn't have to worry about starting the server, deploying the server. We, we, we give it like all the formatting, all like, like when we have random bugs in our platform, we tell it about the bugs, about things that shouldn't do because there are bugs in our platform. And then we give it all of the, um, goodies, all the included platform features. So for example, how to do, uh, open AI with, which you're, which we just saw, we, we explain that it's like available to it and here's how to use it. That's really interesting. Um, and it's nice, it's nice to have that stuff. Cause it's, that's what I would normally have to paste into, um, you know, chat to your cloud or whatever. So it has the, the most update up-to-date docs. Um, I guess like how has the usage been or like, what have, what have people been doing? Like if you're sort of the underlying thing that gets you excited has been how programming changes the way that people think and, um, what, what they can make, like, um, what has been the effect of townie on, on that for you? Yeah, it's been fantastic and also a little bit confusing. I didn't expect to have non-programmers be able to use Valetown so soon. Um, like I I've, that's been a long-term ambition, but, um, there are people who really don't know the first thing about coding and come onto Valetown and make stuff like this in a minute or two and their mind is blown and they're so excited. And that's amazing. And I love that. Um, like, you know, yesterday someone made it like, or recently now that Bitcoin is doing so well, people are making like Bitcoin price tracker, trackers, wallet things, like all sorts of like analyzers for crypto. And like Valetown is really good at that. I could just plug and play with the APIs. Uh, and, and on the one hand, I'm really excited about it. On the other hand, I'm getting questions that are so bad because they're not pro like it's a, a tool for programmers and then I'm getting, and then like, you know, like here, I'll just like the, the one I got last night was this, like, you know, let's just say we remove this and say, save it. And by this, uh, basically what you did is you, is you went into the code and you were, you removed a semi-colon from the, uh, from the JSON object. Um, or I guess it's CSS. I removed like a quote of the CSS. Yeah. I just, I, I purposely put a syntax error in the code, a very simple syntax error that like a programmer would have no problem solving. But if you're not a programmer, you're like, this is the worst experience I've ever had in any software ever. Yeah. I mean, like, what are you going to do with that? If it looks so intimidating, like type error, you know, and if you're a programmer, you're like, okay, I can fix that or whatever. Yeah. I think that makes a lot of sense. That's, that's really interesting. So like, what has that been like for you? Cause I, I've watched you over the last 10 years or so be thinking a lot about the, like the ways to create a programming language that would be powerful, but also, um, simple enough that people can like, people who are non-technical can like easily sort of like onboard and learn to use it. Um, and it seems like a lot of those sort of language design, um, initiatives are just like completely changed by the fact that you can just code with English now. Um, like, what does that, what has that been like for you? Yeah. It's, it's been very surprising. I didn't expect this would be how things are on the one hand. Uh, yeah. Like you kind of rethink things from scratch. Uh, like for example, this type error is totally nonsensical to you and me, but like, look, look at this little button here, like ask townie to fix it. Uh, and then it just sends it to townie and townie figured, immediately figured out the issue. And then like the only crappy part about this experience is that it's going to regenerate all the code from scratch. Uh, but besides that, uh, like, you know, the, the, the example I gave isn't necessarily a condemning thing. Like, you know, they're like the problem for LLMs is just, or the solution to these problems is often just throw more LLMs at it. So I guess given that you like made this for programmers and now it seems like you have like a whole wave of like non-programmers starting to use it. That seems like a big question for like the business of like, who do you want to serve? Um, like, yeah. What are you thinking? I, I'm really tempted to go after people who are non-programmers because it seems like there's just like incredible pent up demand and there's just so many of them and they, they're not served by any other products. And at the same time, I, I really like serving the smartest customers and, and like sophisticated customers and like building a pro tool. It's like, you could think of the difference between like Figma and Canva. Are you going like kind of a tool for everyone or are you building a tool for professionals? I think, I think that's kind of where I'm drawing the line. And like the, the advice I gave to the guy who emailed me about this, this thing right here, uh, like the first day he asked the question, I like solved the problem for him. And in the second day, he asked literally the same exact question. He was like, what's happening? Why do I have this error again? And so my response the second time was, uh, here's how I would solve it. I guess where I'm going at for people, I think there's a whole class of people who are allergic to code. They don't even want to look at it. They don't, they don't even want to know it exists. They totally abstract away from me. And Valentine is not the product for those people. The product for those people is anthropic, um, anthropic artifact, cloud artifacts, I think, or, or GitHub Sparks, maybe. I think there, there are products for people who don't want to look at the code, but Valentine is not that product. Yeah. First of all, I think it's really great that you kind of like have a sense for the kind of customer that you like to serve. Um, I think that's actually quite rare and it's, and even if you have that sense, it's often like quite hard to allow yourself to like be guided by it. Um, but overall, I think you will make a better product and a better company by like following that. Cause like, it sucks to, it sucks to not like your customers or not necessarily even not like them, but like, it sucks to, um, uh, it's much better to really love your customers and want to interact with them and all that kind of stuff. And, and so I think that makes sense. It's funny. Cause I was like, kind of while you're answering, like my response to this is like, and I know very little about your business. So, you know, uh, take this with a grain of salt, but like my feeling is like, I think that there's this new type of programmer emerging, um, which is like, I think a lot of people who are already really great programmers, some of them are using AI, but a lot of them aren't, or they're just like, it's, you know, I can go faster just doing it myself. And there's this like new version of programmers who are like really like AI native and, and programming with AI is like almost a different skillset than programming without it. Um, and right now those people, um, look very unsophisticated, but they will be very, very sophisticated in like 10 years. Um, and so the, like the, the example I would draw is like, if you, if you were like talking to Mr. Beast, like 15 years ago, when he first started making YouTube videos, you'd be like, this guy's dumb, like he's 12 or whatever. Um, but now he's like one of the most sophisticated filmmakers in the world with like some of the biggest budgets. Um, and so my, like my feeling about some of this stuff is like, if you can, if you reach those people now, um, and grow with them and help kind of instill in them some of the like taste that you've built up and experience you built up over many years, like, um, there's a, I think that there's just a chance to make a platform for programming. That's like sort of like YouTube, but it's for programming. It's for building stuff. Um, and those people are not going to be sophisticated programmers. They're just going to be like kids with dreams. I, I would like it to be true that I can, that I can get the future, Mr. Beasts in your analogy who like want to put in the work, uh, to, to me, that's the distinction. Uh, like, I think there are people who want to turn off their brain and like just use an LLM. And I, I, do you know what I'm talking about? Like, I don't want to disparage, because I really do believe with you that there's a new kind of AI engineer who's unbelievably powerful. And, and I love watching those people. And then there's another kind of LLM engineer who will like ask the LLM to do something and then leave their computer, go do something else, and then come back and see an error and just hit the fix it button and then go and leave and do something else and come back and hit the fix it button. And they're like allergic to the code. And I'm, I'm skeptical that those people are going to be, are, are giving themselves the right feedback loops to become the next Mr. Beast. Maybe. Um, I mean, I think like, I'll just speak for myself. Like, sometimes I do that because like, um, one of the things that I think is, is interesting about these tools. And I've talked about this a couple of times before on the show is like, um, right now, programming requires or pre pre now programming requires like a lot of focus and attention. Um, and it still does, but what's really interesting about these tools is, um, you can actually like make progress even when you have fractured attention. So like an example might be like, uh, I have kids and like, I get home from work and I want to work on my side project, but like, I need to like kind of run back and forth between my kids and like the thing I'm working on. Um, or an example is like, I'm a bit, I'm busy. Like I have a lot of work going on today, but like, you know, for me, I have this idea I want to make and like, it's kind of fun to be able to like jump in, like spend five minutes, like fixing a thing and then let it go off and do the thing and come back, you know, and not, that's not to say like, you should get rid of also get rid of focused work. It's just that like, um, programming becomes possible in a fractured attention state, which was it, it was always possible if you had enough money to hire people. Um, but now ever, everyone can do it, which I think is, is kind of cool. And I also think like, I don't know if we extend the YouTube analogy, like to get one Mr. Beast, you had to have like, like a hundred million people uploading, like just random stuff that like they put no thought into. Um, so, so yeah, I think it's, it's, it's, it's kind of interesting. Like, um, I, I totally, I totally see what you mean. Like you want to build a platform for people who like care. And, um, it also seems like this is, we're at such an early stage of this that like, in order to get those people, you have to like, make it so accessible that like tons and tons of people can try it. Um, but maybe I'm, you know, again, you've thought about this way more than I have. Yeah. The, uh, I think a big variable that we're like, not, that we haven't yet talked about is what's happening to the underlying models, like on what timescale. So like Sonnet 3.5 has made all of this, this whole conversation possible. And if we don't have another step change from that in the next year or two, then I think maybe like what I'm talking, like people are going to be more limited by human focus and care. But if we have like another, if like Sonnet 4 is as big as Sonnet 3.5 was relative, then, then what you're talking about makes a lot more sense. Like beginners are, or could get, get even more leverage. So like the rate is, is gonna, but eventually I think we'll get to a point where, um, where, what you're talking about makes a lot more sense. Like you'll just have agents running off in the background. Like they'll be coming up with ideas. They'll be talking to your customers for you, coming up with ideas to make things better. I, who knows? Um, well, I guess, what world are you planning for? Are you planning for a world where like progress is starting to asymptote a bit, or are you planning for a world where we're going to continue to see like the same kind of jumps in progress that we've been seeing? I would like to be like neutral to underlying LLM changes and be prepared to like, if, if the LLM is the same, then, you know, our tool continues to work. And if the LLM gets better, in theory, we just like change the model name and we get the benefits like everybody else did. And I think we saw this happen with cursor and WebSim when three, when Sonic 3.5 came out, they were ready and they jumped. And so I think now Valton is ready for the next model jump. If it happens. That's really interesting. What, um, what have you learned about software development and running software teams in the AI age? Cause it's like very different from, from programming before. And there's like different sets of ideas and like methodologies and it's all, it's all different to have like squishy software. I was talking to, um, Simon last about this and I, I'm curious about like your perspective on it. Like what, what has changed for you? It, for like the, the Val town team, like how we're working with like, like a serious engineering team, serious software versus like software you make in Val town. That's like a lot squishier. Yeah. So right now there's such a huge gulf. Uh, like when you write code in Val town, you have to get so much software configured on your computer and like every week or two, you have to make a tweak to that. And so like, I am the one person on the team who doesn't have that set up on my, like, I can't make that change to our production app. Cause it's just like too much to, to keep that local software up to date with the team. So you make your change locally, you get all your databases, everything running locally. You have to do automated testing. It's like, there's just so much paperwork. It's really important stuff to keep our service running smoothly. Uh, pull, it used to be a pull request. You get like a review on the pull request. You do feedback, you deploy it. Deploying takes 10 minutes. It's, it's like much slower. Uh, yeah, it's a totally different feel than Val town where every save is a deploy. It happened in 50 milliseconds and you're, you're live, you're, you're off into the races. Right. I guess what I'm saying is like, um, doing the programming and ensuring a quality product and running tests and all that kind of stuff is, um, it's significantly different when the code you're running is deterministic versus stochastic. Um, and I'm curious how that has unfolded for you or what you're learning about, like keeping an LLM app running well in production. Got it. Got it. Yeah. It, I get like the word there is evals. Like that's like the, how we as an industry have gotten some amount of predictability or, or understandability about our, how LLMs are performing. We, didn't have real evals for the first three, four months of townie. And we would just like YOLO changes and like, we would like make a change. We tested out ourselves a bunch. We'd deploy it to production. And then like it ended up, it totally seemed better. And then like, we get anecdotal reports that it's worse and we wouldn't really know. And then we would YOLO another change. And it was these things like, I feel like it really brings you face to face in contact with what a, like a truly stochastic thing is like, because like there've been multiple times when I go to Twitter and say like, our users are all reporting X. And like, is it, does anyone else see X? Like, has the model gotten stupider? Has the model gotten smarter? Is the model doing this for other people? Or is it just Valtown? And then, um, like there would be, I feel like I was part of, I don't know if you saw this on, on Twitter. Uh, there was like a couple months ago, this like weird panic where everyone was like, Claude 35 Sonnet has gotten this dumber this weekend. And like everyone in Anthropoc was trying to figure it out. And I think at the end of it all, it was just a panic that people like me helped contribute to. We were, where we all just tried to, we all just kind of convinced each other was super smart and then had a couple of bad it or it like interactions with it. And then there was a panic that it was dumb. It's really hard to reason about these things. And that's why evals was, I feel like evals are important for your mental health. Yeah. That's really interesting. I hadn't even thought about the like social contagion aspect of like the perception of quality, you know? Um, that's wild that that's so, I mean, I guess it's like you can draw analogies to like the stock market. Um, but your product quality being like the stock market is like a very new, it's a new thing. Very new. Yeah. Yeah. And yeah. So like, but evals let, let you kind of like point to something and just like feel good about yourself. Uh, because customers come to me all the time and they like complain about a specific thing. And then I can like look at the evals and see if I can find what they're talking about. And if, and if I can't, I could say like, I'm sorry, try again one more time. It'll probably work, you know, with confidence. Yeah. Yeah. Yeah. Well, I guess then like that sort of runs into the problem of like, then you can only, you only take seriously things that you can like measure, um, or changes that you can measure. Um, like, how do you think about that? Like, you're starting to only see things where there is an eval for it. I think there's a lot we miss. It's like a very squishy product. And I think customers are, our users are constantly having less than optimal experiences and hopefully they're like the tool allows for error correction. Uh, yeah. Like all the time, uh, I'll talk to Jackson, our team who does the LLM, the eval stuff. And he's like, yeah, we don't even have an eval for that. Like, it's not even going to like when, uh, when Claude three, five sonnet new showed up, we were so excited to run evals on it. And like the evals weren't any different because like, we didn't have any evals that were hard enough. Like it would show up that it got smarter. Right, right, right, right. That's interesting. Yeah. I, I definitely, um, I definitely vibe with that. I think like for any product that we've made, um, that I've been the one to start, there are no evals and we just sort of yellowed it for a while. Um, the official term, yellowing it. Yeah. I think that's great. I want to make yellowing happen. Um, um, and I think we've just now started to, um, to add evals like to spiral. And we have a couple of internal incubations that are not released yet that definitely have evals. Um, and, uh, one thing that's, that's been interesting for us is like with spiral, for example, when you like change the model, we also use the new model to change the default prompt that we use to generate new, new things. But then we didn't want to like go and update the prompt for everyone. So we just like have a thing that's like, do you want to upgrade the spiral to like the new version of the prompt and people press upgrade, which has been an interesting thing to try. That's really cool. We've struggled with that internally where if you, we let you edit the system prompt, but then when we push an update to the system prompt, you don't get it. Yeah. Like it's, it's complicated. And, and like, if they've changed it, like, yeah, what do you do? Do you like diff it or whatever? But I think at least like letting people opt in, like letting people know, Hey, there's a new prompt. Um, and letting people opt in is like, uh, it seems to work so far. Um, and then you have, you basically make a copy of the spiral. So you, you don't lose anything. Um, and then maybe you can go back and forth if you want to. Hmm. That's cool. Yeah. It's so you, you put the system prompt as a property of a spiral. Cause right now you're the system prompt in Valentine to the property of your account or like of your settings. Yes. Each spiral has its own prompt basically. Um, cause spiral is like effectively a prompt builder, um, like a fancy, fancy prompt builder. So yeah, each one has its own prompt. One question I have for you is, um, you said previously that, um, what you want to be doing is building an application where you're neutral to the rate of progress. And if the progress is, is to the upside, then it's like a really nice surprise. How do you think about doing that? How do you think about architecting your product and your systems for that? Yeah, it's, it's a good question. And it's interesting because it just happened. Like we built a system for Sonic 3.5 and then Sonic 3.5 new came out and, um, and, and Haiku came out. And in some sense we're like very well positioned for it, you know, like in like beta flags for us internally, we've been playing around with the different models and they just work. Uh, but in, in practice, we haven't actually deployed them to customers yet. We're working on that now. Um, so I think like a model switcher is, is like some pretty good infrastructure. Another thing we're building, we recently got a bunch of new townie users, uh, which has, uh, has made it like extremely important to get pricing, like proper usage based pricing in for townie, because like these things are really expensive and we just haven't taken the time because it's hard to build a whole like new pricing model in place. So like, like those sorts of infra pieces, like being able to switch models, being able to handle multiple models at a time, add new models in kind of more quickly, like you're doing, like when you add a new model, do you have the infrastructure to like tweak the prompt for that model? We, we, we don't like in practice, that hasn't been great. Like Claude 3.5 Sonnet new is doing things, did a lot of bad stuff. Um, and like we had to like tweak the prompt a lot for it. And ultimately we just rolled back to Claude 3.5 Sonnet old because we couldn't get the prompting right for the newer model. Interesting. What about like strategically? Cause there's this like, there's this tension here, right. Where like you're relying on Claude 3.5 Sonnet. Um, but also Claude has artifacts and like artifacts is like, there's some overlap with townie. Um, and same is true for like all these other apps or whatever. Um, and so I think like one of the games of being a startup is, um, that's benefiting from AI is, is sort of strategically thinking about how to, um, benefit from the gains of these companies without also being eaten by them because they all have consumer or B2B like offerings that are not just API offerings. So yeah. How do you think about that? Yeah. Our differentiator has always been that we run back, back end compute. We're like a back end as a service provider. And I think these LM companies don't, won't want to run their own functions as a service infrastructure internally. They're going to want to partner with someone. And that's the outsized dream that they partner with us to help them run back and compute for their customers somehow. Um, yeah, like front end, like even now, like with, with their, the front end apps that they make, like, it's pretty crazy that Anthropic, you know, lets you share the actual front end app, like all, like you can kind of deploy from within Anthropic. Um, but yeah, where, where are these companies going? Yeah, it's unclear, but I think we're like as safe as one could be in back end, in back end functions, service infrastructure. I think that's pretty unique. Yeah. You think that they're not going to build that they're going to partner. And yeah, that's interesting. How are you feeling about things? Like how you've been, you've been in this company, I guess it's been in like a year and a half ish. You just, I think you just, it has been a year and a half. How long have you been running this? Two years. Two years. Uh, you raised around in like February, February, March. Um, yeah. How, how are things, how are things feeling? What is, what has gone how you expected and what is, what has not gone as you expected? And yeah, where are you right now? Yeah. I think things have been great, like have gone better than expected in some departments and worse than others. Uh, we, we just are bringing on like three. So we've been a team of four for like the whole year. And we just hired three more people who are starting in January. Um, so like, I'm kind of sitting on the edge of my seat, like waiting to see how things like the team dynamic shifts. Uh, cause that's like, you know, almost doubling, uh, in terms of growth, like things are growing well, but not like as fast as I want. And it's like interesting in this space to see there, there are a lot of competitors who nobody knows about or talks about and we're like excited to be doing better than them. And then there are crazy success stories that are like taking off to the moon, like overnight. And it's, you know, hard to not be one of those. We're like, we're in the middle, we're doing well. We're having steady progress. Uh, like things aren't going as well as they could. What's an example, like replet or something? Uh, replet isn't what I had on my mind. Um, like bolt on new just announced that they did like $4 million in revenue in their first four months of existence. So that's like, it's like, wow, amazing. So cool. And then, uh, like cursor, I think is another kind of runaway success. Like those are the two that I'd be jealous of, uh, replet agents. I think got a lot of hype, but I don't really know people who use it a lot. Did you use it? Oh, I have not really used it. Honestly. Um, I used replet a lot when I was doing my course. Um, I, like I had this, you know, how to build an AI chatbot course and it was like amazing for that. Um, because being able to like set up a project and like have all the code in the project and students could like fork it and then press run and just works was like, it was amazing. Um, but like for my day to day programming stuff, I'm just, I just use cursor right now. Um, I found that like there were at least back in the back when I was using it, this is before a replet agent, like there was just some frustrating things about it. And like, and the AI was like, not as good. Um, so I, but I don't have an, I don't have an updated opinion on it. Yeah. Yeah. Replet, I think originally and still has product market fit mostly in education. That's like where it shines. So it makes sense that it worked for you in a course setting. Yeah. I mean, I, I know, I know that feeling, um, that kind of like it's going, it's going well, but it's not like blowing up, but it's also not dead and sort of like being, being in the middle can be, it's, I think everyone spends a while in there, you know, it's like, it's called like the trough of sorrow or whatever reason. I don't know if you would call it the trough of sorrow. Cause you are like, things are going well, but it's like, it's, it's, uh, sometimes things going well, but not incredible is like harder than like things not going well at all. Cause you can at least just be like, well, I'm not at least just be straight sad as opposed to being like, yeah, like mostly happy. Um, yeah, I definitely, I feel like, um, I feel like we're in this like, sort of like uptick right now. Um, every or one of your apps, every overall. Um, and I would say like, I would say growth is like, like, like the last couple of months has been really good. This month is like sort of like leveling off a little bit, but I think we'll have a couple more launches in the next couple of weeks that, that will sort of change that. And we're, we're not really doing any like paid marketing or whatever. Um, but I, I, it just feels like things are happening, which is really, really fun. But there were years where it was not like that. So I, I feel you. Um, uh, and, and I, I totally recognize the like sort of team dynamic thing. Like we've, we've grown the team a lot and, um, is this your office? This is not, I mean, this is the office I work at it, but it's not mine. Uh, you know, you know, Jesse B. Rudy, right? No. Um, he's a really good friend of mine. Also went to Penn, uh, I ventures. This is the, he's a partner at IA and this is their office and they graciously let, uh, let us work out of here, crash here. Um, uh, until hopefully one day we'll get an office of our own. Nice. Have you been to our office in downtown Brooklyn? Oh yes. I, I, you walked out for one for once. We, I walked up, we, we actually launched sparkle there. If you remember, um, yes, a very auspicious day. Uh, but we were not able to spend that much time together. Cause I was like running around like, like a chicken with my head cut off. Um, um, but yeah, I mean, things, things, things always break when you add new people and, um, it's always, it's always fun. How have you thought about like, um, team sizing in the kind of AI age? Like, um, how many people you need to hire and who you need to hire and what that says about like your runway and your budget and all that kind of stuff. I haven't really changed my thinking about it. Uh, cause like my role models have always been like Instagram and notion, like a team of on the order of 10, like it's like anywhere from five to 15, uh, even cursors. Like I think like a 15, 20 person team or it was when it exploded. I think like this like small dense, but mighty engineering teams have always been attractive to me. And I, I think AI just makes that more, more tractable for us mere mortals to be able to do that. Uh, but I don't think it, it means that I can hire like four instead of seven. I think I still need the seven, but, uh, like hopefully the four of us will have an easier, sorry, the seven of us will have an easier go of it than without AI. That makes sense. I agree. Like I, um, I feel like the last 10 years, there are the examples of the like Instagrams and notions, a huge team now, but like they, they were pretty small for a while. Notion was like six people when they did like database, like in 2019, when I was using notion when we were all using notion, like it was a real company band. It was like six people, something crazy. Yeah. Yeah. Yeah. Um, but, but I, I do think like, but there was this whole like movement for like big teams and whatever. And, um, I, I really love having a small team. I think it's super fun. Um, and everyone knows each other and everyone's like friends and like, and it just, it changes things when it's like 40 or 50 or a hundred, a thousand people in a way that's like, um, yeah, it's like less personal, but obviously like you can get less done, but now depending on what you're doing, like you can actually get really far with smaller, smaller team, which sounds just more fun to me. Yeah. I don't think a small team is holding, you know, software is one of those things, uh, more people don't make it happen any faster than mythical man's month and everything like, uh, having like really tight context, like everyone on the team knows what's going on. Iterate really fast. Cause like that's what we're doing right now, for example, is adding, like creating like a virtual file systems kind of set up. So right now, Valtown is so good at writing single file apps, but then you want to break it up into a couple of different files and have the AI just like edit one of them for you while the other ones they put, but Valtown can't handle that. It's just like, we haven't built that infra, but now we're building it and there's no, like there's no parallelizability of it. Like they're like, they're just like core infra decisions. Like the whole team has to kind of be on board with, and we can't add more people to make it go faster. More people would honestly make it go slower. Right, right, right. How do you think about like, once you're, now you're getting into multifile edits, like, how do you think about what the appropriate unit of work is for the AI to do before it comes back and is like, I did this. And what does that imply for like, how much in the loop should you be versus not? Yeah. I think for better or for worse, we've set ourselves up to be spectators in this race. And like, uh, we like watch cursor and anthropic and, you know, bolt and whoever, like come up with these new UI interaction patterns and AI interaction patterns. And then we take the best ones and we like adopt them. I think that that's what's worked best for us to be like caught fast copycatters. Cause like our core business is, and like what we want to be best at and like innovate on is like a functions as a service platform. That's like the simplest and fastest, like most pleasurable to use. And then we like layer on these AI features that like other people will innovate for us. Uh, so that's interesting. Yeah. It's, it's weird to be like, cause I feel like my instinct is like, no, no, no, we have to, we have to be the best. We, we're going to come up with this stuff, but the industry is just moving so fast. If we just wait a month or two, we can see what other people do and copy it. That's what we did with Townie. Like Townie was just like, we, we, we, we looked at, um, before anthropic artifacts, Townie was like a tool use thing that looked just like chat CPT at the time. We were like, just kind of copying chat CPT and it was pretty bad. And then anthropic artifacts launched and we're like, ah, there we go. That's, that's the correct UI pattern. And now we're like on that, but now we're watching, you know, cursor invent all these new UI patterns. We're like, ah, yes, those are the ones we want now. That's interesting. Um, does it like, how does it sit with you? Like, and, and how do you stick to your street? Like you had, it seems like you have a very crisp idea of your strategy of like what you want to be best at, right. Which is like, um, backend as a service functions, um, is what you said. And that like strategy was formulated, um, even before all the stuff came out or was like really popular. Um, and it seems like you're getting this pull with Townie from people who like, you know, are maybe not even programmers and maybe don't really understand what backend as a service even is. Um, and so how are you dealing with like the like excitement of that? And, but also you're able to say like, that's, that's important. Um, but we're going to be sort of fast, fast followers because like the core strategy is this other thing. Like, yeah. How are you thinking about that? It's you really hit the nail on the head. It's, um, uh, in the last week or two, I've had a lot of discussions with our main investor, Dan Levine from Excel. He he's like a spiritual co-founder for this company. Uh, and like a lot of the core strategy pieces about like what we are and what we aren't come from him. And we had a meeting last week where I was like kind of showing him like this alternative vision, like we could pivot from the backend as a strategy to a like full stack app development tool. Like maybe you're like YouTube for programmers vision or like Shopify kind of a, like a full stack app platform, uh, like Shopify for SAS apps. You could, you is like one idea, like we could rebrand, we could call ourselves SASify or something. Um, and like go after like bolt and like try and get like hundreds of thousands of non-programmers to just be like buying credits and credits and credits, uh, to like make, make apps. Like I think that's a real pivot we could plausibly do. Um, and we can have a database included and like all sorts of other services included. And then we become probably a tool more like Canva where we're like for everyone we're for, we're, we're, we're for everyone and like unsophisticated folks. And they're like making apps, you know, like, so, you know, kind of like small, medium sized apps, like no serious engineer would, would use us. Um, and again, I had a long chat about it and the various trade-offs, uh, we had a long chat and, uh, I wrote up this like long strategy document about what our strategy is, what it, and very importantly, what it isn't and why we, why we're giving it up and like how we get pigeonholed if we do the other strategy. Like one of the big downsides of being like a full stack app is that you have to do the whole stack. And I think it seems simple and it's always seems simple and it's just so much more complicated than everybody thinks in order to do front end and back end and monitoring and errors and logging and databases and database migrations and database backups. Just like we're talking about like a mature, back to your question from earlier, a mature engineering organization has like a dozen or two dozen tools in order to, including like a local text editor and like local version control GitHub, like so many, so many tools. And so if I want to replace that whole stack of like two dozen tools, uh, it just necessarily has to be a toy version of all of them and like a kind of bad version of all of them. And then that customers necessarily become unsophisticated. It, it, it leads, leads to a weird local maxima that we're scared of. Uh, and then the alternative is that we do fewer things, but do them really sharply and better and integrate with all those other things. Like we, we could become just the functions of service piece for a lot of people. And then we can, once that's kind of nailed, we can expand more confidently while remaining quite good at each thing we do instead of mediocre at all of them. That makes a ton of sense. What, um, what makes you think that that's like the part of the stack that you want to sit in? Like why choose that one? Why backend as a service, uh, or yeah, why functions as a service? It's a really, really good question. Like in some sense, the most honest answer is that, uh, it's, it's what like Dan Levine thought there was an opportunity for. It's like what our investors had as an opportunity for, or it's like what was left, uh, like Dan's also an investor in Vercel and they're doing front end and they're like doing such a good job at the front end. Uh, and then there's like a backend shape opportunity. That's, that's like, that's one way to look at it. Uh, I've always like another way to look at it is I've like, as long as I've been a programmer, I've really wanted something like this to exist. I've like always hated how complicated backend stuff is. And, uh, like there are like three dozen tools that I've used over the years, trying to make my life easier on the backend and never found one that was good enough. Uh, and, and so like wanting it to exist, like nobody, well, I guess like we could, we could talk about who else is in this market. There's like eight, there's AWS Lambda, which nobody would ever use directly. Like you have to use it like intermediate to all sorts of stuff. It's like a terrible, terrible user experience compared to normal programming. And then there's a cloud floor workers, which is actually a wonderful product. Uh, and, and I've learned a lot from, from the cloud floor team, but still it's, it's, it's not social. It's not GitHub. It's not like collaborative open source, productive coding like GitHub is like when you make a cloud floor worker, it's, it's kind of invisible code that nobody else can then leverage and use and like, and for it's not a, so it's not an artifact and it's not a browser-based thing. It's a thing you develop like a normal engineering artifact. Yeah. It's like, um, it's sort of a last generation type thing. Um, like for like people working in more enterprise environments where there's like less flexibility and there's, you know, like, yeah, I think, I think that that makes sense. Um, I, I, I just gotta say, like, I think it's awesome that, um, that you're thinking through these things like so crisply and you have a really like clean perspective on it. Um, which, uh, I think it's awesome one. Cause it like, um, it's just, that's just rare. And to like, you're giving yourself the ability to be wrong, which also gives you the ability to be right. Um, whereas I think it's like, there's a tendency to like try to do everything and be everything says the man who like has the studio where you have like five different products or whatever. It's literally called every, trying to do everything. Um, so, uh, so I, I think I admire that because like, um, I know internally how hard that is for me. Um, and I, I just really appreciate, um, like getting to hear it and that you're willing to share it. Like, I think a lot of people are not, not willing to share it. Um, I think you're just being incredibly honest and I think you have a very clear vision for what you're doing and I'm just like psyched to see what you do with it. Like what happens in the next year or two. Thank you. Yeah. All right. Me too. I think it's the vision's been the same since day one, but it's remarkable how many twists and turns there are even staying true to an original vision. Totally. Um, well, this is awesome. I had a great time. I'm really glad you came on the show. We got to do it more often. Um, if people want to find you and find Val town, where can they find you on the internet? Uh, man, I, I normally say Twitter, but I'm, I'm trying to move off of it these days, uh, vaguely, but yeah, I'm, I'm Steve Krause, uh, kind of everywhere on the internet and Val that town is where you can find Val town. Awesome. Thanks Steve. Thanks Dan. Oh my gosh, folks. You absolutely positively have to smash that like button and subscribe to AI and I. Why? Because this show is the epitome of awesomeness. It's like finding a treasure chest in your backyard, but instead of gold, it's filled with pure unadulterated knowledge bombs about chat GPT. Every episode is a roller coaster of emotions, insights, and laughter that will leave you on the edge of your seat, craving for more. It's not just a show. It's a journey into the future with Dan Shipper as the captain of the spaceship. So do yourself a favor, hit like, smash subscribe, and strap in for the ride of your life. And now without any further ado, let me just say, Dan, I'm absolutely hopelessly in love with you.