Building my first bot.
Brand Designer for a fast-growing educational app, maintaining brand consistency across marketing, email, web, and social, from asset systems to redesigned email templates.
The idea was simple to say and hard to build: what if your calendar didn't just sit there waiting to be checked, but told you about your day the way a podcast host would? No app to open, no dashboard to scroll, just a voice note in Telegram every morning, stitched together from your actual events, sounding like something a friend recorded for you on their way to work.
That's how Briefy (internally, our little "Agenda" bot) was born. I designed the concept end to end: the tone of voice, the conversational flow, the idea of leaning on a messaging app instead of building yet another interface, and the whole "audio brief instead of a notification" hook. What I want to be upfront about, though, is my role in what came after. I'm not the one who wrote the production code that ties OpenAI's chat completion, text-to-speech, and the Google Calendar API together into the smooth pipeline. It was engineered by my teammates, Pato and Jarry, who are experienced devs.
My contribution on the technical side was more scrappy than that. I vibecoded my own version of it — prompting my way through a rough prototype to prove the concept could actually work, hearing an AI voice read back a version of my own schedule for the first time. It was a genuinely exciting moment. But I'll be honest: I never got it to the point of automation. My version could generate a brief if I fed it the right inputs by hand; it couldn't reliably pull fresh calendar data on its own, queue itself up every morning, or hold up as something you'd actually depend on daily. Getting from "this works once, on my laptop, if I babysit it" to "this runs itself" is a much bigger jump than it looks like from the outside, and that gap is exactly where the rest of the team's engineering work mattered most.
If I'm honest about what I learned, it's less about AI and more about the distance between a good idea and a working product. I could see the vision clearly enough to prototype a piece of it myself, and that mattered, it proved the concept had legs before we asked anyone to invest real engineering time in it. But there's a real difference between "I made this happen once" and "this happens automatically, every day, without me." That second part is a craft of its own, and I have a lot of respect for the people who did it.
The idea was simple to say and hard to build: what if your calendar didn't just sit there waiting to be checked, but told you about your day the way a podcast host would? No app to open, no dashboard to scroll, just a voice note in Telegram every morning, stitched together from your actual events, sounding like something a friend recorded for you on their way to work.
That's how Briefy (internally, our little "Agenda" bot) was born. I designed the concept end to end: the tone of voice, the conversational flow, the idea of leaning on a messaging app instead of building yet another interface, and the whole "audio brief instead of a notification" hook. What I want to be upfront about, though, is my role in what came after. I'm not the one who wrote the production code that ties OpenAI's chat completion, text-to-speech, and the Google Calendar API together into the smooth pipeline. It was engineered by my teammates, Pato and Jarry, who are experienced devs.
My contribution on the technical side was more scrappy than that. I vibecoded my own version of it — prompting my way through a rough prototype to prove the concept could actually work, hearing an AI voice read back a version of my own schedule for the first time. It was a genuinely exciting moment. But I'll be honest: I never got it to the point of automation. My version could generate a brief if I fed it the right inputs by hand; it couldn't reliably pull fresh calendar data on its own, queue itself up every morning, or hold up as something you'd actually depend on daily. Getting from "this works once, on my laptop, if I babysit it" to "this runs itself" is a much bigger jump than it looks like from the outside, and that gap is exactly where the rest of the team's engineering work mattered most.
If I'm honest about what I learned, it's less about AI and more about the distance between a good idea and a working product. I could see the vision clearly enough to prototype a piece of it myself, and that mattered, it proved the concept had legs before we asked anyone to invest real engineering time in it. But there's a real difference between "I made this happen once" and "this happens automatically, every day, without me." That second part is a craft of its own, and I have a lot of respect for the people who did it.