Most AI learning content is passive. You watch a video, you read a thread, you feel informed for about 48 hours, and then nothing sticks because you never actually practiced the technique.
The AI Advent Challenge was designed around one principle: the only way to learn an AI skill is to use it, every day, on a real problem.
900 participants spent 25 days in December doing exactly that.
The Problem With How People Learn AI
There is a structural problem with AI education right now. Content is abundant, but practice is rare.
Courses give you theory. Tutorials give you copy-paste prompts. Threads give you a hot take. None of them require you to actually sit down, apply a technique to your own work, and experience the difference it makes.
The gap is not information. It is daily, hands-on practice. People know they should be using AI better. They do not have a structure that makes practicing feel achievable every single day.
The AI Advent Challenge was built to close that gap: 25 sequential lessons, one per day, each one a practical AI technique you could apply immediately to whatever you were already working on.
The Stack
The AI Advent Challenge runs on:
✅ Lovable — product development and interface
✅ Supabase — database, auth, and state management
✅ CampaignDays.io — campaign and event infrastructure
✅ TypeScript — application logic
No agency. No dev team. Built and shipped solo in under 30 days before the December launch window.
The PRD Decisions That Defined the Product
Sequential Unlocking Was Not a UX Choice. It Was a Retention Decision.
The most important architectural decision I made was to lock each day’s prompt behind the previous one. You cannot skip to Day 12. You cannot binge the whole thing in one afternoon.
That constraint was deliberate, and it created the daily return behavior that makes the product work. When you build a learning product, the enemy is not low-quality content. It is the human tendency to consume everything at once and then feel done without having learned anything.
Sequential unlocking made each day feel like a commitment rather than a choice. The 42% day-one activation rate tells me the structure was compelling enough to get people to start. The fact that it runs annually tells me it is compelling enough for people to come back.
How I Measured the AI Advent Challenge with Pirate Metrics
The first version of my analytics dashboard showed one number prominently: task completion rate. It looked great. It told me almost nothing useful.
The Contributor Model Was Unplanned and Became the Real Product
I launched the platform and then put out an open call to the community. I had no expectation of the response I got.
37 AI experts submitted prompts. Each one chose a specific technique they had personally tested, refined, and used in their own work. Their reasoning, their edge cases, their real-world applications came with the submission.
What I thought was a content sourcing shortcut became the actual product. 37 different practitioners, 37 different professional contexts, 37 different proof-of-application moments. Community-sourced knowledge organized as a structured learning path is more valuable than any curriculum I could have designed alone.
The contributor model was not in the original PRD. It is now the core product differentiator.
Shipping in Under 30 Days Was a Constraint That Made the Product Better
I used Lovable to build the entire platform in under 30 days. That timeline was not comfortable, but it was the right call.
Building fast against a hard deadline forced me to make sharp scoping decisions. Every feature that did not directly serve the “one lesson per day, sequential unlock” core got cut. The result is a focused product that does exactly what it promises, with no bloat.
The architecture decision that made this possible: treating each “door” as a simple state object in Supabase rather than a complex content management system. Simple data model, clear unlock logic, reliable delivery.
The Architecture That Made 900 Participants Show Up Every Day
The hardest problem in a 25-day challenge is not day one. Day one is easy because it is new. The hard problem is day eight, when the novelty is gone and there are seventeen days left.
What the Numbers Look Like
900+ participants in the December run
42% day-one activation rate from registered participants
37 community contributors who submitted original prompts
Built and shipped solo in under 30 days
Runs annually, with the contributor model now a core fixture
The activation rate is the number I am most proud of. Getting 42% of registered users to complete Day 1 in a product that requires daily commitment is not typical for event-based learning experiences. The sequential structure is doing its job.
The Untold Story
37 AI experts each chose one technique that changed how they work with AI. Their reasoning, their experiments, their lessons, all organized into a 25-day path through real applied AI practice.
That is a story worth telling properly. More details coming soon.
Who the AI Advent Challenge Is For
✅ PMs and operators who use AI daily but want to use it more intentionally
✅ Professionals who have read about prompting techniques but never practiced them systematically
✅ Anyone who wants a structured, daily habit for improving their AI workflow
✅ Anyone who has absorbed a lot of AI content and wants to actually apply it
The challenge runs in December. Participation is free. The techniques are permanent.
Join the Next Run
The AI Advent Challenge runs every December. If you want to be the first to know when registration opens, get on the list at aiadventchallenge.com.
If you are building an AI learning product and want a product audit before your next launch, book here at promptledproduct.com/audits.