Proof of Vibes
An approach to building confidence and exploring the creative possibilities of AI tools: Gather your people. Learn through making. Share your experience.
What began as a way to create a little structure, direction and accountability rapidly unfurled into a group expedition for six AI-curious product makers. Design friends who I’d been increasingly chatting with about AI’s tremendous possibilities and terrifying potential pitfalls.
The brief was simple: Ship a project. On your own. In 4 days.
We’d meet every day to check in on progress, plans and reflections. We’d share our demos and learnings at the end.
Essentially, a hack week, themed around using AI to help us all gain a more immediate understanding of how these tools will change and accelerate how we work.
AI is everywhere
We can read the hot takes and chase every new tool, but nothing helps you understand this stuff better than building something yourself.
Yes, there’s Ghiblification and action figure characters, and that former coworker who vibe-coded that impressive project, but how does this really impact your own process?
As the week unfolded, I quickly realised two truths:
AI isn’t quite there yet. I lost hours caught in loops just trying to name folders in Xcode. All the hype I’d seen online suddenly felt very far away.
But when it works, it really amplifies existing skills. With a clear idea of what you want to build, and a bit of persistence, it can massively accelerate your workflow.
The key to all of this is having a clear vision of where you’re trying to get to and consistent trial and error to iterate toward the future, spotting the endless loops as they emerge and unblocking yourself to move past them.
Scope and iteration
The project I chose was an app to deepen the connection between a parent and their children through a series of short prompts that are randomly chosen when you open the app.
Because it was simple I thought this a good case study to try making an iOS and Apple Watch app.
I’m no software engineer but I’ve written code in the past and am comfortable getting my hands a little dirty.
But I’d never even opened Xcode before.
I figured if AI tools really were the magical superpower I hoped they were, this would just be a small bump in the road. But after spending most of the first two days running into frustration after frustration in Xcode, I hit a wall. That was when I turned back to the group, and their tools, shortcuts and suggestions helped me gain fresh perspective and change my direction. Switching to a simpler web stack wasn’t just a technical change. It was a shift in mindset, sparked by hearing how others were working through their own roadblocks.
From here my progress accelerated massively. Replit was my tool of choice, enabling me to write as much JavaScript in 30 minutes as would have taken me a day or two before. I loaded in a CSV, selected a random row, loaded the data, added basic animations etc. Nothing super complex, but I understood it all.
The hare and the tortoise
AI’s promise encouraged me to start with large ambition and nothing more than a prompt. As I pecked away at the keyboard, sharpening my prompt as the real-time feedback wasn’t getting me exactly what I wanted, I started to realise the missing building blocks weren’t actually words after all. They were things like:
A simple colour palette and typography system.
A wireframe or diagram to show how different parts of the experience should fit together.
A simple illustration for style and placement.
While AI can do remarkable things with a simple natural language prompt, the real value is in finding the fastest route to clearly communicating your ideas. As someone with a design background, this is often far quicker by drawing or making something in Figma that AI could then help bring to life.
In our group reflections I kept coming back to the old fable of the hare and the tortoise. While I’d entered this week expecting to be the hare, sprinting straight to the finish line, I actually found most progress when I approached problems as the tortoise, taking the time to clarify simple design artefacts before AI could rapidly accelerate things from there. Perhaps a tortoise with a jetpack is a better analogy.
Learning together, as we went
The accountability and structure were wonderful. I created far more space to explore AI tools than I ever would have done on my own. But the real value came when I got stuck. Seeing how others pushed through their challenges, whether by changing tools, rethinking their prompts or simply talking through the problem, helped me look at my own blocks differently. I stopped seeing them as dead ends, but as inspiration to explore new avenues.
My favourite two:
Speech-to-text. Tom inspired us all to try speaking to our computers as a method of accelerating typing speed, and getting more into the flow of truly ‘vibe-coding’. While this took a little getting used to there, I can definitely see the potential. But not in the way I expected. AI works best when it has a ton of context, beyond the specific request you’re making. So being able to regularly brain dump everything on your mind in relation to an instruction you’re about to give seemed to significantly improve the chance of this leading to the sort of outcome you had in mind, even if it feels quite haphazard in giving this direction.
Just use Replit. Noam suggested using Replit. Others agreed. On the spectrum of vibe-coding tools it’s toward the more accessible end, but it was incredibly quick to get something live at a URL I could share with people. At this point, since I was building a website, it was also very easy to step in and make manual edits when I needed to. The complexity hadn’t been abstracted away too much.
But overall, there was the feeling of figuring this out together, poking holes and pushing AI into unfamiliar territory with others who’d gain from that experimentation too.
What we built
And of course, here are the projects we built.
Digestible News Web App. Chris built a web app that pulls articles from the NY Times and Guardian APIs, summarising them into digestible bullet-point cards using AI.
Tiny Sparks. I built a web app for encouraging thoughtful conversations between parents and children, with AI-generated prompts and character illustrations. Replit Link.
Pocket-to-Podcast Chrome Extension. Tom built a Chrome extension that saves articles, uses OpenAI to generate a podcast script, and Eleven Labs to create an audio version, hosted via Buzzsprout. Chrome store link
Mighty Veg. Noam built a mobile-optimised Veggie Recipe Converter web app that takes photos of meat-based recipes and suggests vegetarian substitutions, complete with AI-generated ingredient illustrations.
React Native Birthday App. Jonny built a React Native app that connects to an API to store birthdays, notifies users, and features tab navigation and styling with AI-assisted token design.
HealthApp Screenshot Library. Liam built a FlutterFlow-based web app to upload, categorise, and tag mobile app screenshots with AI-generated descriptions and filters.
What’s next?
Will AI change everything? Probably. But the only way to find out what that means for you is to roll up your sleeves and try.
Don’t wait for the perfect tool or the perfect moment. Pick an idea. Set a timer. Gather your crew. See what you can ship in a week.
You’ll be surprised how much further you go when you’re not going it alone.
Big shout out to Gary Chou’s Gather Your People, which has long inspired me to try a group format like this.
LOVE THIS! been doing this on my own and not with as much structure, so this felt timely yet again!
Did u folks try lovable/v0? Would u suggest starting with an iOS app or web app if I were to try?