Archived

Inspiration

If you're in the right circles, you find out over coffee that your startup idea has already failed three other companies. If you aren't, you spend a year and a lot of energy learning the exact same lesson. All those startup failure data are sitting online, but they're just static lists. Nobody looks at them when they first have a breakthrough idea. We built this so that archive could finally talk back.

What it does

Drop in your idea, and we match it against a graveyard of dead startups. For every match, the system generates a post mortem. It will find the proximate cause they told the press, the root cause that actually killed them, and the change so that your idea does not fall in the same traps. We also have a forum for communities to come and share their experiences, question, and common pain points.

How we built it

We designed the UX to be highly tactile and immediately actionable, avoiding standard dashboard templates entirely. The interface is literally a desktop OS in the browser: you type your idea into a Spotlight bar, and the 173 dead startups sit in a Trash Bin where they belong. Under the hood, the matching engine is built to never fail on stage. We committed the embedding weights directly into the repo, so a live demo never hangs on terrible conference WiFi. When you search, it checks a Postgres cache first for instant results. If it’s a new query, a LLM streams a strictly constrained report. If the API goes down, it gracefully falls back to raw hard data.

Challenges we ran into

Getting the data ready was a massive task. Here is what we actually had to do behind the scenes:

Cleaning the mess: We started with 815 names but narrowed it down to 409 that actually had useful details. We had to manually fix broken text formats and garbled characters.

Forcing accuracy: We ruthlessly cut unverified data. If a startup's failure only had one source backing it up, we dropped it.

UI/UX: We were brainstorming on what would be the appearance of the app itself.