A Conversation About the Future of Creative Technology

A dialogue between two builders exploring the overlooked frontiers of artificial intelligence

Jack: You know what strikes me about most AI discussions today? Everyone’s talking about the same things: chatbots, image generators, code assistants. But there’s this entire landscape of possibilities that barely anyone’s exploring.

Arman: I’ve noticed that too. It’s like we’re all crowded around the same few applications while these massive territories remain completely untouched. Where do you think we should start?

The Broken Philosophy of Productivity Software

Jack: Let’s start with something that surprised me: the fundamental brokenness of productivity software. Not the AI part; the entire philosophy behind how we organize our lives. Think about every productivity app you’ve used: Notion, ClickUp, Asana. What do they all have in common?

Arman: They’re all about structure. Organization. Breaking down tasks into manageable pieces.

Jack: Exactly. But here’s what I’ve observed: most people don’t fail at productivity because they can’t organize their tasks. They fail because of fear, overwhelm, and loss of motivation. You can spend fifteen minutes every morning planning your day in Notion, build this elaborate system with the perfect database properties and views, and within two weeks? You’ve abandoned it completely.

Arman: The classic Notion graveyard. I’ve built so many systems that I used for maybe a week.

Jack: The reason isn’t that the system was bad. It’s that maintaining the system becomes another source of stress. There’s no emotional support built in. The software doesn’t understand why you’re avoiding that important project. It doesn’t know that you’re procrastinating because you’re afraid of failure, not because you forgot about the deadline. Imagine productivity software that starts with a simple question: Why do you want to achieve this goal? Not just “what are the steps?” but “what’s driving you?” The AI becomes a companion that understands your motivation and adapts to how you’re feeling each day. Some days you need a push. Other days you need permission to rest. The software should know the difference.

Arman: That’s a completely different paradigm. Instead of organizing tasks, you’re organizing emotional energy.

Jack: And here’s the thing: everyone has ambition. But keeping yourself consistent for three months? That’s where people fall off. It’s easy to do something for a week. The real question is: can we design software that helps people push beyond that one-week mark? Take making friends as an example. You can’t measure that with “I’ve met five people this week, goal accomplished.” Friendships are qualitative. Some are shallow, some are deep. A simple progress bar is meaningless. You need an AI that can actually understand the quality of your progress by analyzing your reflections and experiences over time.

Arman: So the AI would track subtle signals rather than just checkboxes. But how would you make this engaging enough that people actually stick with it?

Jack: Here’s where it gets interesting: narrative. People love stories. What if your productivity journey became an RPG? Your real-world goals are quests in a genre you choose: sci-fi, fantasy, whatever. You’re generating your own hero’s journey, but the quests are actual things in your life. Going to the gym becomes a training montage. Studying becomes leveling up your skills. And maybe there’s a social layer too: imagine Duolingo streaks, but for your real life goals. You and your friends can see each other’s progress, nudge each other when someone’s falling behind. Not in a guilt-inducing way, but in a genuinely supportive way.

Arman: Gamification, but personalized and narrative-driven rather than just points and badges. The existing gamification is shallow. Real games work because they have story, progression, meaning. Speaking of things nobody’s really solved: music generation fascinates me. There are tools that can generate music, but something’s missing.

The Music Generation Problem

Jack: The fundamental problem is balancing AI generation with human creativity. It’s easy to have AI generate a complete song. It’s also easy to give humans complete control over every parameter. But the interesting question is: how do you find the middle ground where both you and the AI have meaningfully contributed? That’s the question that could be a startup in itself. Let me give you an example. With beat production, one of the most time-consuming parts is finding the right drum sample. You literally go through hundreds of kicks, snares, and hi-hats, playing each one until you find what works. What if AI could understand the “vibe” you’re going for from a description and surface the right sounds?

Arman: So AI as assistant rather than replacement. It handles the tedious search, but you make the creative decisions.

Jack: Exactly. But here’s where it gets really interesting: voice synthesis. Vocaloid works by recording a person singing every single pitch and syllable, then stitching them together. The result sounds distinctly robotic. There are abrupt endings on each note, no natural modulation between pitches. What would natural synthesis require? Smooth transitions between notes. Dynamic expression: vibrato, breathiness, the ability to belt or whisper. The capability to incorporate effects like opera-style pitch changes or even beatboxing. Essentially, the synthesis needs to understand singing as a continuous, expressive act rather than discrete phonemes.

Arman: We discovered Synthesizer V during our research. That seemed impressive.

Jack: Incredibly impressive. They have a million users, raised $20 million, generating an estimated $10 million yearly. And we’d never even heard of them. But here’s what nobody’s built yet: a complete music composition system where you have AI agents handling different aspects (melody, drums, vocals, mixing) but you maintain fine-grained control over each element. The mixing is the hard part. It’s like unscrambling an egg. Going from forty separate tracks to one mixed song is straightforward; you just combine them. But taking one finished song and separating it back into forty tracks? Nearly impossible.

Arman: So the individual pieces exist: lyrics generation, melody creation, voice synthesis. The innovation is in the integration. But there’s another challenge: how does an AI know if music sounds good? You can’t just train on popular music, because popularity doesn’t equal quality. I’ve noticed that with underground artists. Some of the most interesting music I discover is completely different from the mainstream, and that uniqueness is exactly what makes it successful.

Jack: So maybe the answer is personalization: fine-tuning models on your music taste rather than trying to optimize for universal appeal. Generate music that you would like, not music that theoretically everyone should like. And here’s something else nobody talks about: most music production already follows fairly universal patterns. The music theory is consistent, the typical workflow is consistent (drums, then bass, then synths, then melody, then vocals). We could create AI agents that follow that proven workflow while giving users control points at each stage.

Arman: So respecting how musicians actually work rather than trying to replace the entire process. Let me shift to something that came up when we saw those AI-generated games. The technology is impressive, but I keep thinking about how games could be more personally meaningful.

Memory-Aware Personal Games

Arman: Imagine a game that has access to your message history: all your texts, chats, conversations. It analyzes who you talk to, what you do together, where you spend time. Then it generates quests based on your actual relationships and activities.

Jack: So instead of generic objectives, it’s “play basketball with Devon” because the AI knows you play basketball with Devon regularly.

Arman: Right. But here’s where it gets emotionally interesting: maybe there’s someone you haven’t messaged in a year, but you used to be close. The game generates a quest like “get lunch with Alex.” You complete that quest in the game, reliving that connection virtually, and then maybe, just maybe, you’re inclined to actually reach out to Alex in real life. The game becomes a mirror for your relationships, prompting you to maintain connections you’ve let slip.

Jack: There’s another layer we could add here: real geographic data. What if the game knew the actual coordinates of places in your life (your gym, your favorite restaurant, your school, your friend’s house)? Then it could generate a world with accurate spatial relationships and realistic travel times. The AI would need to handle ambiguity (there might be multiple Korean barbecue places, but it could use context to figure out which one you mean). And once it knows where you are, it can create realistic scenarios. Driving to that restaurant takes twenty-five minutes and costs two dollars in gas. You spend money eating there, but you build relationship points with whoever you’re with.

Arman: You’ve essentially created a real-life simulator. And we could extend it further: want to meet Elon Musk? The game could take real information about Elon Musk and generate a narrative quest chain that might plausibly lead to meeting him, or tell you it’s genuinely too difficult and suggest more achievable aspirations.

Jack: Right. Your game world contains your real friends and real places, but it also incorporates your ambitions and role models. It becomes both a reflection of your current life and an exploration of possible futures. The technical architecture for this would be node-based. Each location is a node. Characters are nodes. Events are nodes. Everything can connect to everything else through relationships. You start with known entities from the message history, but the world can expand dynamically: when the story needs a new location, you generate it with appropriate context that maintains consistency with what exists.

Arman: So it’s structured but infinite. Bounded but expandable.

Jack: Exactly. And here’s the crucial part: we’d start text-based. Not because we can’t do graphics, but because text lets us focus on the core innovation (the memory integration, the quest generation, the relationship tracking). Build that foundation solidly, then layer on visuals later if it makes sense. That’s actually more innovative anyway. We discovered that when we tested existing AI game generators: they use templates for the mechanics and graphics, then AI generates the content and story. The template provides the proven framework; the AI provides the creativity and customization.

Why Isn’t Anyone Building This?

Arman: You’re not generating everything from scratch each time. You’re using structure to enable creativity rather than constraining it. So we have three significant ideas here: emotionally intelligent productivity software, integrated music composition tools, and memory-aware personal games. The technology exists to build all of them. So why isn’t anyone doing it?

Jack: I think part of it is that people are crowded around the obvious applications. Everyone’s building chatbots and image generators because those are the examples that went viral. There’s also the complexity factor. These aren’t simple wrapper apps around an API. They require thoughtful integration of multiple AI systems, careful UX design, and deep thinking about what the technology should actually do for humans. The music project is a good example. Individual pieces exist (lyrics generation is straightforward, melody models are available, voice synthesis is advancing rapidly). But combining them coherently? That’s genuinely hard. It’s not just a technical challenge but a design challenge.

Arman: And with the productivity software, there’s a philosophical hurdle. The entire industry is built around the assumption that better organization leads to better outcomes. Challenging that assumption and saying “actually, the problem is emotional, not organizational” requires rethinking everything.

Jack: The game idea has a different barrier: privacy and data access. Building something that analyzes your personal messages requires extraordinary trust and careful handling of sensitive information. But I think the deeper reason is that building something genuinely novel requires resisting the urge to copy what’s already successful. It’s much easier to build “ChatGPT but for X” than to ask fundamental questions about what problems AI should solve.

Arman: And most people underestimate what they’re capable of building. They see polished products from well-funded companies and assume that level of sophistication is unreachable.

Jack: Which is exactly wrong. We tested a game generation platform during our research, and you know what we found? It’s not magic. They use templates for game mechanics, then AI generates custom content, quests, and assets. It’s clever engineering, not impossible technology. The components are accessible. What matters is how you combine them and what problem you’re solving.

Where to Start

Arman: Let’s say someone reading this gets inspired. Where should they actually start?

Jack: Start with the problem that personally frustrates you. For me, it was the singing voice synthesis. I experienced that pain directly when creating music videos. For you, it might be something else. But genuine frustration is the best starting point because you’ll have intuition about what a good solution feels like. And scope it appropriately. We kept talking about MVPs: minimum viable versions that prove the concept without requiring years of development. For the music project, the MVP isn’t a complete composition suite. It’s: take lyrics, take a melody, synthesize natural singing. Three components working together well is more valuable than ten components working poorly.

Arman: For the game, we’d start text-based rather than graphical. Build the memory integration and quest generation first, prove that’s compelling, then add visual layers later if needed.

Jack: And for the productivity software, maybe the MVP is just the AI companion that understands your motivation and creates bite-sized quests. Skip the social features, skip the RPG narrative, skip the sophisticated progress tracking. One thing done exceptionally well. This is the discipline that’s hard for people: accepting that you can’t build everything at once. But it’s also liberating. You can actually finish something instead of drowning in ambition.

Arman: Looking across these three ideas (productivity, music, and games), is there a common thread?

Jack: They’re all about human experience rather than pure capability. The productivity software isn’t about organizing tasks more efficiently; it’s about understanding human motivation. The music tools aren’t about generating any music; they’re about helping people express themselves musically. The game isn’t about graphics or mechanics; it’s about reflecting and strengthening real relationships. So the pattern is: use AI’s capabilities in service of human needs that aren’t purely technical.

Arman: Right. And they all involve integration: combining multiple AI systems thoughtfully rather than just using one model for everything. There’s also something about personalization. The productivity software adapts to your emotional state. The music generator learns your taste. The game is built from your actual life. These aren’t one-size-fits-all solutions.

Jack: Which makes them harder to build and scale, but also more valuable to the people who use them. And I think that’s the opportunity. The big AI companies are optimizing for the broadest possible use cases. But there’s enormous value in applications that go deep on specific human needs, even if they’re more complex to build.

Are We in an AI Bubble?

Arman: There’s something I’ve been wondering. Do you think we’re in an AI bubble?

Jack: I think we’re in a bubble, but the bubble might actually be real. It means that yes, a lot of things are overvalued. A lot of AI applications are overhyped. There’s definitely froth and speculation. But the actual impact and potential warrants at least 50% of the current excitement. Maybe more. Think about it this way: can you imagine life without ChatGPT now? I know there’s a charm to coding without AI assistance (I appreciate that). But AI has become so indispensable in so many workflows that going back would feel like going from smartphone to flip phone.

Arman: That’s exactly how I think about it. The nostalgic feeling I get remembering flip phones versus smartphones? I’m starting to get that same feeling about the pre-AI era in high school and middle school. There is genuine charm in figuring everything out yourself, in looking through multiple websites filled with ads and bloat to find a recipe. But would I actually choose that over asking an AI and getting a clean answer immediately? No. I’ll take the efficiency.

Jack: So we’re in a moment where the technology is genuinely transformative, even if the hype cycle is ahead of actual deployment. The interesting question isn’t whether AI matters (it clearly does). The question is what we build with it. While everyone’s building obvious things or just wrappers around existing models, the opportunities exist for people willing to think differently and tackle harder problems.

---

Jack and Arman are builders exploring the intersection of AI and human experience. They believe the most interesting applications aren’t the ones everyone’s talking about; they’re the ones nobody’s built yet.