Introduction
Welcome to our latest developer interview, where we dive into the world of AI applications and gain insights from the creators themselves. Today, we’re talking with the developer behind Pico, an innovative AI chat app designed to make writing and revising text easier, even for non-native English speakers. Initially created as a personal tool, Pico has grown into a versatile app with multiple personas for various writing tasks and tight OS integration.
Product
Developer
Ronald Mannak
Twitter Handle
@ronaldmannak and @picoGPT
Can you share a brief overview of your AI app and the problem it aims to solve?
English is my second language and I used to spend a lot of time on writing emails and text. And I never knew whether I used the right tone. I originally created Pico for myself, to save time. I would give Pico bullet points and have Pico write an email or text in any style I wanted.
While testing early versions of Pico on TestFlight, I learned I wasn’t the only one with these issues. A lot of people, even native English speakers, found Pico to be useful. That’s when I decided to focus Pico on writing and revising text. Later I received many other requests for things Pico could do, so over time, Pico became a general AI chat app.
What sets your AI app apart from others in the same domain?
We’re at the very early stages of AI. I quickly learned people were having problems writing good prompts. While some users loved experimenting and creating mind-blowing prompts, most people just wanted the AI to work. At that point I decided to include multiple personas to Pico, each with a useful or fun task. The main persona is focused on writing emails. There is a persona that writes tweet storms. There is even teacher persona is interactive and teaches you any subject you want. I’m regularly adding new personas, based on user requests.
Another unique feature is OS integration. For example, Pico supercharges Siri. Responses can be sent via Gmail or Apple Mail in a single click. And on Mac, Pico is available in any app via a right mouse click. Whether you’re using Google Docs, Apple Pages, MS Word or a completely different app, Pico is always just a (right) mouse click away.
What sparked your interest in AI, and how did you begin your journey as an AI app developer?
I first started experimenting with GPT, Dall-E and Stable Diffusion during the Holidays last year (2022). I was in Canada visiting my girlfriend’s family, and every day I showed the family a new prototype. It was fun to see people being so enthusiastic. I knew right away this is something I needed to work full-time on.
What challenges do/did you face during the development process, and how did you overcome them?
I revised the UI a dozen times. While Pico looks very basic and simple, it took a long time to make it that simple.
But the most time I probably spend on AWS, creating a proxy server to the OpenAI API. When it worked, it was great, but sometimes calls would seemingly fail randomly. It took me a while to figure out calls always failed after 30 seconds. That was a breakthrough and using Google I found out AWS has a hardcoded 30 second timeout. I don’t know who thought that would be a good idea, but for AI applications it sure doesn’t work. Note I still used Google at that point of the development process. I hardly use Google anymore. In fact, I just asked Pico why my calls to AWS fail after 30 seconds, and Pico gave me the correct answer faster than I found the answer using Google.
How do you envision the impact of your AI app on users’ lives and the broader community?
I integrating more iOS and macOS features into Pico. I want Pico to become an AI layer in between the OS and applications. Every app should be “AI enhanced”. WWDC is in two weeks. We’ll know then whether Apple jumped on the LLM train or dropped the ball. I see huge opportunities either way.
In the long term, I see LLMs becoming the core, the UI layer and the app layer of operating systems. Basically, the LLM will become the OS. Apps and App Stores will disappear since you will be able to describe the app you want to the LLM. All apps do just manipulating data. If the LLM knows the data format, it is able to manipulate the data. It doesn’t matter to the LLM if the data is a PDF, a PNG or accessible through an API. I don’t think many people are aware how our world is going to change in a few years.
What future updates or developments do you have planned for your AI app?
Lots of fun stuff coming out soon. More personas, even better OS integration and improved visual representation. I have more ideas than time to implement those ideas.
Can you share one piece of advice for aspiring AI app developers?
Nobody know what is going to happen in the next few months, let alone in the next few years. I have never seen a technology evolve so fast as AI. Every week there’s a major breakthrough. Every day I see great new ideas.
I’d say: just build and experiment as much as you can. It’s the most fascinating time to be alive IMO.
Conclusion
It’s clear that AI app developers like the one behind Pico are shaping the future of technology and how we interact with it. With a focus on continuous improvement and integration with existing systems, Pico aims to become an AI layer between the OS and applications, ultimately leading to a world where apps and app stores are replaced by AI-driven systems. Aspiring AI app developers should seize the opportunity to experiment and build in this rapidly evolving field. Stay tuned for more developer interviews as we continue to explore the fascinating world of AI applications.