Designers gain valuable insights from projects, but they often get lost in notes, forgotten in docs, or stuck in memory. When starting new work, recalling past decisions is hard, and turning insights into something useful, like a case study or social post, feels like extra work.
I wanted to build something that makes it easy to capture, find, and repurpose those learnings so they don’t go to waste.
Role:
Product Builder
Duration:
48 Hours
Tools:
Figma
Bolt.new
N8N
Supabase
Claude.ai
ChatGPT
Outcomes:
Information Architecture
Systems Mapping
Functional Web App
AI Agent Knowledge-base (RAG)
Context:
I participated in a 48-hour weekend sprint to rapidly design and build a product idea I had been thinking about for a while.
It was a fast-paced, self-driven hackathon, and my goal was to create something that I could present to people for feedback, which would help me improve the product.
"Weekend Hackathon 2 hr update:
It took me a little time to wrap my head around how everything should work, but I finally have a rough idea of what I want to build. I also put together some diagrams to map out how these functions will work on the backend.
The goal is to let designers log their learnings, pull them up later when working on similar projects, and even turn them into stories, whether that’s a social post, case study, or blog. I’ve sketched out the main flows, including how AI will help retrieve insights and recognize key moments.
Next step: getting the RAG setup up and running.
Sharing some early system diagrams—would love to hear what you think. What’s one feature you’d want in a tool like this"
"Alright, wrapping up Day 1 of the Weekend Hackathon, and I’ve got a basic front-end prototype up.
Getting to this point was a bit chaotic at first; I started building directly in Bolt, but without a clear reference, things got messy real quick. So, I took a step back and threw together a basic wireframe in Figma, which made a huge difference. Once I had a structure, handing it off to Bolt felt way smoother, and I could actually see things take shape without second-guessing every layout decision. Also, I was saving so much on the token usage, so that was a plus!
Now, I have:
- A basic login screen.
- A note editor where users can type in their learnings.
- A chat interface that will eventually let users pull up past insights dynamically.
- A basic projects page where all projects will show up
Tomorrow’s focus:
1. Setting up the database and hooking up the backend to save data
2. Implementing the AI assistant to make learnings searchable and reusable
3. Building the storytelling aspect of the product.
Feeling good about the momentum so far. If you were using a tool like this, what’s one thing that would make it really useful for you? Let me know!"
"Kicking off Day 2 with some solid progress; user login and database storage are now working!
Yesterday was all about getting the basic front-end in place, but today I shifted focus to setting up the database using Supabase. Now, users can log their learnings and assign a project to each learning. This data is then stored in a Supabase database in a vectorized format. The data for each user is unique to them and is ready to be used with an RAG engine to power the chat feature.
It’s still early, but now that the core structure is in place, the next step is making it useful. That means:
1. Building the AI Agent and setting up RAG for the chat functionality.
2. Leveraging this vector data store to help designers with storytelling.
3. Prompt Engineering the AI functions to generate accurate results.
Feeling good about where things are headed! If you’ve ever tried to revisit past project insights, what was the biggest pain point? Would love to hear what would make this most useful."
"This is where things get exciting—I’ve set up an AI-powered chat that lets you talk to your own past learnings!
With n8n handling the backend, Supabase for storage, and OpenAI to make sense out of it, the system can now:
- Retrieve insights from your personal knowledge base
- Understand context and surface relevant past decisions
- Help connect the dots between different projects
This means if a designer wants to recall why they made a certain decision on a past project, they can just ask instead of digging through old notes. The AI fetches the most relevant insights and provides responses tailored to what they already know. (Ofcourse, this is still a prototype and may not be 100% accurate)
Next, with the time I have left, I will focus on:
1. The project page UI.
2. Telling a story from all your logs.
3. Prompt Engineering the backend prompts to get better results.
This is feeling like something I’d personally use every day. If you could have an AI-powered assistant that remembers all your past work, how would you use it?"
"And that’s a wrap on this 48-hour hackathon!
After two intense days, I’ve touched upon 3 key aspects of the product:
- Log their learnings and capture insights as they happen.
- Chat with their own knowledge base wherein AI retrieves past decisions with context.
- Tell a story about their work with AI-generate narratives.
I’ll be honest, I had moments of doubt throughout this. Would this actually be useful? Can it really fit into a designer’s workflow? I don’t know if this works yet. And that’s okay.
But that’s exactly why I’m putting this in front of people. The best way to improve a product is to test assumptions, see what resonates, and iterate.
Some things didn’t make it into the 48-hour build:
- Growth insights: Spotting trends and patterns across projects
- Integrations: Pulling in insights from Notion, Slack, and Figma
I’ll be refining this and making it available for designers to try. If you’re curious or think this could be useful, let’s talk! Would love to hear what would make this actually work in a real workflow.
48 hours, a lot of doubt, but even more learning. Now, it’s time to see what happens next. How do you test your product ideas before fully committing?"