doombase.io

Inspiration

The idea for DoomBase.io came to me the night before the hackathon began. I was thinking about how much noise and anxiety surrounds AI news today — and how nobody really knows where we’re headed. I wondered: What if we had a dashboard for existential risk? Not a boring academic tool, but something strange, emotional, and even a little fun — like a panic meter with commentary from fictional AI personalities.

The next morning, I started building.

The problem? I had never written a single line of Python prior to a few weeks earlier when I needed a quote tool for my concrete batch plant. It all started there. Also, I work 55–60 hours a week, which didn’t leave much time. I didn’t know FastAPI, Supabase, or how to deploy a modern React app. Pretty sure I still struggle with all of those things.

But I knew I had to see this idea through — even if it melted my brain.

Through the vibe-coding process of putting all this together, I learned a lot. And that, in essence, was the point. Not to win anything — just to learn.

DoomBase.io curates and scores real-time AI doom signals — Reddit posts, YouTube videos, Substack essays — and presents them in an animated Bolt.new dashboard with commentary from fictional AI personas.

What it does

It includes:

A crawler + editor + analyzer pipeline
Tone rewrites and severity scoring by custom AI agents
Persona commentary from EliezerBot, EviBot, RAWBot, BostromBot, and HALBot
Supabase-backed database storing curated doom entries, comments, and top headlines
A cinematic UI featuring a panic index, terminal interface, ticker bar, and summary cards
If the doom score spikes too high, HALBot takes over the UI. (This is not installed yet — but coming.)

How we built it

Backend: Python + FastAPI to handle crawlers, curation, and agent pipelines
AI Curation: Ashby (rewriting editor) and Aneliezer (doom analyzer)
Persona Agents: Custom GPT-based personalities with unique logic and tone
Database: Supabase for all entries, tags, mood logs, top headlines, and feed
Frontend: Built in Bolt.new and deployed via Netlify
Automation: n8n orchestrates scheduled crawls and planned social media post-processing
Crawlers: yt-dlp, PRAW, SerpAPI, and custom Substack scrapers

All wired together into a fully automated doom aggregation system.

Challenges we ran into

Just learning what all these terms meant: vibe coding, Bolt, Python, n8n, MCP, React, Tailwind, Supabase, etc.
Learning Python through pure experimentation — and wiring up FastAPI, Supabase, and frontend deployment
Managing UUID consistency across Supabase insertions (still my nemesis)
Building five AI agents with personality, weighted logic, and fallback options
Debugging database insert flows, frontend data pulls, and timing issues
Automating workflows with n8n, cron jobs, timestamps, and error handling
Staying awake long enough to finish after 60-hour workweeks

Accomplishments that we're proud of

A working end-to-end AI curation system, written in under 30 days
A fully functioning live frontend built in Bolt.new
The ability to pull Reddit, YouTube, Substack, and SerpAPI doom entries automatically
Persona agents that feel alive and respond to the world’s doom score
Creating something weird, beautiful, and surprisingly useful — with zero prior experience

Even if everything doesn’t work exactly as intended yet, I’m proud of what does — and what’s coming. There’s heart here, wrapped in the quiet discomfort of a world on the edge.

It’s deadly serious news baked into a satirical lemon cream cheese birthday cake with sprinkles.

What we learned

How to build a full-stack AI product from scratch — even under time pressure
That storytelling, tone, and humor can make even complex systems feel human
How to glue together APIs, databases, and AI agents into a real product
That the best ideas sometimes come from chaos and curiosity (though I’ve always known that — this was just the best proof yet)

What's next for doombase.io

Add user accounts, Doom Journals, panic streaks, user badges, and gamification
Launch a “Time Capsule” feature — users can write future predictions and reveal them a year later
Turn persona agents into contributors for the DoomBase blog, Distotech — with in-universe interviews
Expand crawlers to track government disclosures, AGI research, and scientific papers
Cover fringe trends like robotheology, AI-human relationships, and nation-state alignment responses
Build educational tools, user-curated entries, and a public Doom Digest email
Automate social media video output
Release limited-edition HALBot UI takeovers based on real-world events
A full-blown archive and data tracker where doom scores and social sentiment evolve over time — hard data that might just tell us where we are in history