Making Prvctice, Again

The path to making an app such as Prvctice (Practice) was a couple of months or years in the making, depending on how you look at it. Prvctice was the name of a small studio I started in 2015. At the time, I was in need of a pivot. People had an expectation of me based on the success of my first app, Letter to Jane, a digital magazine app that had a small moment back in the day. I loved everything that opened up for me because of that, but I felt as if I needed a new place to explore without the pressure of others' expectations. Through that process, a new app was made: Dumb Fun, an AR collage app for iOS, as well as other experiments in VR and AR. As I moved on with life, the brand became dormant. I ended Dumb Fun to focus on other projects, and I mainly kept the URL out of nostalgia.

This version of Prvctice would be born out of another need to explore the changes in technology at a time when I again was looking to pivot. At the start of 2024, I set out to understand AI for myself. Most of the marketing was so off-putting, but the technology was so intriguing that I wanted to try to learn what was possible. I started out small with some bespoke code projects and, with the help of my friend Derrick Schultz, began building my own workflows in ComfyUI. That experience was eye-opening, but I didn't feel like I had the right ideas at the time to take advantage.

Also, in 2024, I started writing a book on film theory. One of the exciting parts about working on the book was the research phase, my favorite hobby. My process can best be described as no visible process whatsoever. It mainly involves me writing down very loose, train-of-thought-type ideas while I read or watch something. It can be a word or a phrase, just whatever I feel in the moment. I then come back to these piles of notes, create a bulleted list, and start to dive into each bullet. After a couple of months, I had hundreds of pages of notes, and with a full-time job, I had very little time to process and organize all these thoughts. One day, I decided to pour all of it into ChatGPT just to see what it might do, and it came up with 8–10 high-level themes pulled from all the notes. Half of the ideas weren’t great, but the other half were some of the major points I was looking to write the book about. That was the moment I knew I had found something of value.

Like a lot of people who try an AI chatbot, I had an initial “Wow!” moment followed by a lot of disappointment. Asking the apps about something I didn’t know produced fairly mediocre results. Topics that don’t appeal to general audiences and aren’t based in science and math would only produce surface-level responses. Asking the apps to help with creative projects felt like working with a camp counselor doing group projects more than anything serious. It was clear that any engineering gone into creative thinking was geared toward inspiring a general public more than helping a professional, (I also want to stress that I don’t disagree with that strategy while these are in such an early phase of development). This is where I started to wonder if there was a path for me to make something for myself. I didn’t initially think about coding my own app. I looked into other solutions, but at the end of the day, none of them allowed me to explore in the ways I wanted. Not only did I want to make a research tool, but I also saw this as an opportunity to explore new types of interactions and design workflows. So I dusted off my 2015 MacBook, the same MacBook I used to make Dumb Fun, and started to see what was possible. While choosing to use a 10-year-old computer to build an AI app sounds weird, it was a very intentional choice for a couple of reasons.

AI has some potentially conflicting discussions going on. On one side, AI is all about the need for more power, more compute. We need new chips, new computers, new devices, etc. You can see it in all the commercials for Apple Intelligence or a Microsoft Co-Pilot ad. The other side of the discussion is how AI can potentially benefit all of humanity. If it is a technology that is supposed to work for everyone, then it needs to be able to work everywhere, not just on the latest and greatest hardware. If you log in to X/Twitter, you will see many users addicted to rapid-release schedules. Every week, a new bell and whistle appears that is pushed as a signal of the future. I started to feel more comfortable as I’d been in these discussions before.

I got my degree in photography at a time when the medium rapidly changing. Red cameras had just come out, and everyone was theorizing what an image was. Almost twenty years later, we can all agree that while photography has greatly transformed, we still relatively know what an image is (at least, I hope). After college, I found myself working in publishing. I started out making a print magazine, which I quickly pivoted to digital because the costs were too great. I became one of the faces of this new trend in publishing, largely for just offering an easier experience than a lot of major publishers were pushing at the time.

For those who remember what early digital magazines were, they were a mess. You couldn’t select or copy the text; video was forced into every page regardless if it was needed or not. Overcomplicated animations and sounds turned the reading experience into more or less bloated adware. With Letter to Jane, I offered an experience that was simple. I believed that we didn’t need to reinvent the book or the reading experience, or at least not overnight.

The greatest design advancements come organically. They come from someone sharing an idea and letting others make it their own and learning from that. Time and time again, I’ve witnessed that when you overload people with too many ideas at once, the general response is for them to stick with what they know. You have to start somewhere, and you should start with your best idea and try to make that as good as you can. If you can start there then it’s much easier to build new ideas on top of that. Contemplating all of these things on a five-hour bus ride back from upstate New York after Thanksgiving, I started to finally have a product form in my head. I still had the name Prvctice, and it probably made the most sense for a product like this than it ever did for my brand. I wanted to make something that non-tech people could understand. This meant clearly defined actions, minimal design, and a clear use case. Being intentional in every decision not only helps clarify the product, it’s also economically necessary.

In general, I think we talk a lot about the energy usage of something like ChatGPT, but we conveniently ignore our energy usage in streaming media and games all day, web searches, and refreshing social media. It’s a conversation we need to have, but we need to have it on a larger scale. In general, I believe we need to start asking ourselves how we use the internet and at what cost. We have so many duplicative tools these days, and whether you believe it or not, building this app is an attempt at me being more mindful of my own digital footprint. For the public version of the app, I’m using OpenAI’s Assistant API and GPT-4o mini. I researched other services, but at the moment, this was the most cost-effective for me to balance functionality and speed. I also chose not to include generative media capabilities at the moment because I felt there were better tools and methods out there. No need to waste money and energy on bad images.

Another consideration to go the assistant route was so I could integrate other functionality. The basic design is that the assistant is there to handle general questions while using open sources such as Wikipedia and YouTube to pull information instead of using its own knowledge. This helps safeguard against hallucinations and prevents it from becoming a walled garden. A large part of the "secret sauce" to this comes in the assistant’s programming as well. I’ve taught the AI a general theoretical framework that I like to work in, using concepts from Guy DeBord, Serge Daney, Roland Barthes, Susan Sontag, and more. I also created a custom knowledge base designed to help the assistant look deeper than most web searches or AI apps when it comes to creative conversations. The knowledge base down votes a lot of popular art and artists, not because I don’t like those artists, but because neural networks try to give what they think will be the most generally accepted answer which creates surface level recommendations.

I also chose not to make the assistant have a human-like personality like a lot of apps. One of the worst things about all the big chat apps is that when you start talking about creativity, they start to sound like the most pretentious person you’ve ever heard in a coffee shop. I’ve banned a lot of overly flowery language, but it’s still nowhere near perfect. The balance I’ve found is letting the assistant be curious instead of all-knowing. When I originally had the assistant as a superintelligent professor, it almost exclusively gave me wrong information in the most annoying way possible. The more I simplified its personality and asked it to be more curious, the more I started to feel like the assistant was helpful.

I also wanted to extend this idea into the animations. I may change my mind later, but not having the assistant’s answer "stream" like most apps was meant to also reinforce that this is a tool more than a digital friend. The “thinking” is inspired by old anime. I also love how these old shows presented sci-fi tech as having an organic feel to them. Making the app background akin to a sort of digital pool helped reframe the user's relationship to the app. Again, it’s not about a superintelligent person trapped beneath the UI but instead the feeling, for me, of sending my question into this dream-like space, looking for an answer. These digital waves are also interactive, allowing users to tap and make “splashes” in the waves, which for me is more enjoyable than watching the assistant type out the response.

The last major design choice came from a lesson I learned recently when I was working on the social app Spill. Themes, or “colorways” as they were called, were something I very much wanted as a way to make the app feel more personalized. In a world of endless feeds, there’s very little room for a user to make the app feel like their own. I spent hours and hours coming up with what I felt were these super sophisticated and fun color combinations. Overwhelmingly, people usually hated those. Users by far preferred the neon pink and purple colors that I would’ve never initially thought of doing. While this might seem small, it was a fairly big lightbulb moment for me because it showed how much demand there was to break the monotony of app design in tech. People love to curate their lives in every way they can, and so far, AI has not addressed this. If people are going to feel more comfortable using something new, it should be able to come to them, not the other way around.

From Thanksgiving to early January, I now have a preview build to share. It’s been the fastest development cycle of my life, and that is more a state of where technology is going than because of my clear vision. This is an app purposely made by feel. This is not meant to be a mass-consumer product (that would be awful; I’m paying for this all out of pocket, ha). I have a lot more ideas of where I’d like to take this, but I feel as if I’m at the point where I need to live with what I've built for a while and tweak some things as I go. There will be plenty of time to figure out the next tools to integrate.

The next phase will be building out the private version of this app. Some plans for it are allowing me to choose which model for specific use cases (there’s a lot of Gemini and Anthropic tools I’d like to try out). I want to add more functionality, such as controlling smart home devices and searching other databases and resources. I want to build out the voice capabilities more, but I am waiting for the cost to come down on that. Overall, this has been an amazing experience so far. It’s felt great to get back into coding while also building something that has added real value to my life. Just in the last few weeks of testing, I’ve been able to find new interviews, clips, and documentaries that I had never heard of. I’ve gotten great book recommendations that have already helped me strengthen my research and my writing.

Prvctice is not the most advanced or the flashiest, but I hope it is one of the more useful and sensible tools that I can build. In my mind, the app is at about 20% finished, and as I’m developing a roadmap, I look forward to this evolving and growing as it becomes more integrated into how I do my research. It’s also the first time in years I’ve had a platform to experiment freely. I have no plans at the moment to release this publicly other than to some friends and colleagues to try. If you have read this and would like to talk more, feel free to reach out. I am very open to this being a collaborative process or finding ways to share more things behind the hood.

–Tim Moore, 01/06/2025