Point Party began as a weekend experiment to test the limits of “vibe coding” — the idea that you can build real software by simply chatting with an AI. What started as an experiment turned into a production agile estimation tool now used by multiple squads across Skyscanner.
Vibe coding is basically a means of creating software via conversational AI. Just as you might discuss a feature idea with teammates, you describe the software you want in plain language and the AI turns it into working software, inferring requirements and writing code. You’re not writing specs or coding - you’re just having a conversation like you would in ChatGPT.
I tried both Bolt and Lovable, two of the most established vibe coding tools, and which sell the vision of making software-building accessible to everyone. I ended up mostly using Bolt - partly because I hadn't used it before and wanted to try it out, but mainly because I found it's integration with Supabase (the backend/database for your software) seemed to work a bit smoother for what I needed.
My teams had been using Hatjitsu for some time and though fine, I felt it fell short certain areas. For example, the virtual rooms would often close unexpectedly so we'd have to create a new one, disrupting our refinement; or as a mobile team with platform-specific tickets I'd like to see who has/hasn't voted in a given round.
So I built ‘Point Party’ which is an estimation tool for teams. It helps teams run collaborative refinement sessions in a shared virtual room much more effectively.
Here are the main features:
See who’s voted with names (or hasn’t).
Story point stats - see the average, median and a handy distribution chart.
Anonymous mode, where names are hidden and shuffled.
Host controls - reveal votes, reset rounds, or hand over the host role. Important when you need to dip or if the wifi drops.
Nudges to remind people who haven’t voted.
Multiple estimation scales: Fibonacci, T-shirt sizing, linear, etc.
You can try Point Party out for yourself below. Any feedback, please reach out to me on LinkedIn.
Building Point Party was a fun and useful experience in learning how to vibe code in practice, and build software for which I personally had a painpoint that I wanted to solve. The biggest takeaway was that with these tools the conversation becomes the spec. Every prompt, every tweak, every bug fix is essentially performing an accelerated version of the product development flow. It's quite addictive to be able to shorten the cycle between an idea and build, and go through a much more rapid iteration loop.
I build Point Party in June 2025, and despite being in their infancy these AI vibe coding tools are already really impressive (and no doubt will get even better). I was impressed with the quality of the work from a UX stand point and in some of the product decisions it made to fill in the gaps - e.g. I didn't have to specify anything other than 'virtual room' and it made the leap to have unique room IDs and means of creating/sharing/joining that room in the product.
It made me realise how much speed changes the equation. When you can go from concept to a working product in a few hours, suddenly it’s worth building things that normally wouldn’t make the roadmap — internal tools, one-off ideas, experiments for small user groups, more prototypes. It lowers the bar for what’s “worth shipping.
And then there’s the iteration loop which again is super easy. You can add a feature, test it, roll it back, or push further — all inside one ongoing chat. That speed of feedback makes the process feel less like writing specs and more like a live product jam session. That's why since building Point Party, I've also been using Lovable and Bolt to build more internal prototypes for upcoming Skyscanner features I'm working on - where in a few hours I can go down various paths of what a given product might look like to hone in on a much more polished idea which I can then take to design and engineering to start the more formal processes.
The final lesson? Vibe coding is a great tool in the product toolkit - it can be used for prototyping, and in some niche scenarios for production software. With the right prompts and product instincts, you can build things that hold up in real-world use. It certainly does not replace engineering — but it definitely changes how fast product teams can move ideas from thought to something more tangible.
See who has/hasn't voted 💯
Get stats on estimations 📊
Take over / hand over host controls 👑
Anonymous mode🕵️
Nudge teammates who’ve drifted off 💤
Strikes when your estimates all match 🎯