Quick answer: Global Fail Map is a free, open-source research app that maps failed companies, cancelled megaprojects, abandoned inventions, terminated science programs, and futures that never arrived. It ships with 200 curated stories, custom map markers for different failure categories, sourced Markdown reports, random discovery, and live Valyu DeepResearch for new places or ideas. Try it or fork it.
We built Global Fail Map because we wanted a better way to study ambitious failures.
This is a startup-postmortem theater and not "10 lessons from Theranos". This is not a motivational thread pretending every dead company died from one neat cause.
I wanted the map. Failure is easier to understand when it has coordinates and real people behind it.
What does Global Fail Map do?
Global Fail Map gives you a globe covered in documented attempts that did not reach their intended future.
The first version ships with 200 curated stories across five categories:
- Companies
- Megaprojects
- Science
- Technology
- Visions
Each category has its own marker symbol and companies do not look like bridges. Science does not look like planned cities. Clusters get their own symbol too, because a group of seven stories should not pretend to be one random case.
Click a marker and you get the short version. Open the report and you get the longer version with citations.
The point is simple: if someone spent years and millions, or billions, discovering that a plan did not work, we should not throw away the evidence.
Why this is useful
Most builders study winners all their lives. Understandable but winners are easier to sell, to write about, and to compress into a clean story.
If you are actually building something, the failures are often more useful.
Concorde is not just a plane story. It is a unit economics story.
Fordlandia is not just an odd Henry Ford story. It is a lesson in ignoring local systems.
Theranos is not just fraud. It is what happens when verification becomes socially expensive inside an organization.
The Texas supercollider, the Superconducting Super Collider, is not just a cancelled science project. It is a reminder that science still needs political durability.
Juicero is a hardware margin, distribution, and necessity problem wearing a very expensive shell.
Global Fail Map shows you what broke, where, and what survived?
Who should use it?
Use Global Fail Map if you are:
- A founder studying why companies die before you copy a playbook from the winners
- A developer interested in dead platforms, abandoned hardware, and weird infrastructure
- A journalist looking for sourced starting points instead of recycled summaries
- A student or teacher looking for better case studies
- A researcher studying technology, infrastructure, science, planning, or economic history
- A policy analyst trying to understand how public projects get stuck
- A curious person who enjoys falling into very expensive rabbit holes
You do not need to be a coder but if you do code, the repo is there.
What happens when you click a marker?
Every atlas entry has structured metadata:
- Title
- Subtitle
- Location
- Country
- Coordinates
- Category
- Status
- Period
- Summary
- Lesson
- Report path
- Source links
The report goes deeper. It explains what was attempted, why it mattered, what changed, what evidence supports the ending, what is uncertain, and what remained afterwards.
Some failures leave ruins, court records, and patents. Some leave safety rules and successor projects, while some leave nothing but a domain name and a bankruptcy docket.
The Five Categories
Companies: dissolved companies, bankrupt firms, discontinued corporate programs, dead platforms, and business models that did not survive contact with the market.
Megaprojects: cancelled railways, unfinished airports, ghost metros, dams, ports, bridges, energy projects, new capitals, and public works that got trapped between money, politics, land, and engineering.
Science: terminated trials, withdrawn claims, medical failures, research scandals, and large scientific programs that ended because evidence, safety, cost, or governance broke the plan.
Technology: abandoned inventions, discontinued devices, dead computing platforms, failed vehicle programs, unreleased products, and technical bets that did not become the future.
Visions: utopian settlements, planned cities, charter city proposals, speculative districts, and grand designs that looked better in the model than in the world.
Why the map markers matter
I did not want a generic pin map. Generic pins make every story feel the same. A failed airline, a cancelled dam, a dead clinical trial, and an abandoned planned city need different visual language.
So the app uses custom marker symbols. A company marker reads differently from an infrastructure marker. Science, technology, and vision cases have their own shapes. Dense areas group into clusters so you can explore without visual noise.
The marker system is small, but it changes the feeling of the app. It turns a list of failures into an atlas.
The part I love the most is: the Random story Feature
Search is there when you know what you want. Random story is there when you don't want to type and just explore fast!
Press it and the app throws you into the archive. One click might land on a failed airship company. The next might land on a cancelled nuclear plant, a ghost metro, a dead social network, or a futuristic city that never became more than a promise.
This is the feature that makes the app feel less like a database and more like a museum with trapdoors.
The research mode
The pre-built 200 stories are not the whole idea. The app also gives you the ability to run on live deepresearch.
Search for a place, project, company, or idea. Choose the effort level. Let it run. The app tracks progress and gives you a sourced report when it is done.
Examples:
- "Cancelled smart-city projects in Africa"
- "Failed electric vehicle startups in California"
- "Abandoned nuclear projects in Eastern Europe"
- "Dead social networks in Asia"
- "Unrealized capital city projects"
That makes Global Fail Map both an archive and a research workflow.
How this is different from a chatbot
Sometimes a chatbot is the right tool. If you need a fast explanation of Concorde, ask a chatbot.
But a chatbot does not give you a globe. It does not let you browse by marker type and also does not keep a curated atlas of 200 sourced cases. It does not treat location as part of the evidence.
Here is the comparison:
| Need | Chatbot | Global Fail Map |
|---|---|---|
| Quick explanation | Yes | Yes |
| Geographic browsing | No | Yes |
| Category-specific markers | No | Yes |
| Curated failure archive | No | Yes |
| Source-backed reports | Sometimes | Yes |
| Random discovery | Not really | Yes |
| Fresh deep research | Sometimes | Yes, through Valyu |
Use a chatbot for orientation. Use Global Fail Map when the geography and evidence matter.
Tech stack
The app is built with:
- Next.js 15 App Router
- React 19
- Tailwind CSS
- Mapbox GL JS with globe projection
- Zustand
- react-markdown and remark-gfm
- zod
- Valyu DeepResearch API
- Vercel Analytics
The bundled data lives in src/data/examples.json. Cached reports live in public/reports.
The curation pipeline lives under scripts/ and includes validation, draft generation, and merge scripts.
Running it locally
Clone the repo:
git clone https://github.com/yorkeccak/global-fail-map.git
cd global-fail-map
pnpm install
Create an environment file:
cp .env.example .env.local
Set the basics:
NEXT_PUBLIC_APP_MODE=self-hosted
NEXT_PUBLIC_APP_URL=http://localhost:3000
NEXT_PUBLIC_MAPBOX_ACCESS_TOKEN=pk_your_public_token
VALYU_API_KEY=valyu_your_api_key
Run it:
pnpm dev
Open http://localhost:3000.
How the bundled reports work
Each bundled report is a cached Markdown file. The app does not need to regenerate those on every page load.
The story metadata and source list live in src/data/examples.json. The readable report body lives under public/reports.
There is also a curation workflow for adding more cases safely:
npx tsx scripts/check-curated.ts scripts/curated/africa.json
npx tsx scripts/build-entries.ts
npx tsx scripts/merge-entries.ts
The merge step guards against duplicate IDs, malformed entries, missing coordinates, weak lessons, and missing sources.
This matters because a map like this gets bad quickly if the data discipline slips.
Good First stories to open
If you want to understand the range of the app, try these:
- Fordlandia
- Concorde
- Theranos
- Juicero
- Project Cybersyn
- The Texas supercollider, the Superconducting Super Collider
- N-1 lunar rocket
- Ryugyong Hotel
- New Harmony
- The Line
They are different kinds of failure, which is the point.
Some are technical. Some are political. Some are financial. Some are cultural. Some are evidence failures. Some are coordination failures. Some were bad ideas. Some were good ideas wrapped in bad assumptions.
Frequently asked questions
What is Global Fail Map?
Global Fail Map is an open-source research atlas that maps failed companies, cancelled megaprojects, abandoned inventions, terminated science programs, and unrealized visions. Each marker links to a sourced story and report.
Is it open source?
Yes. The repository is on GitHub.
What is included in the first release?
The bundled atlas includes 200 curated stories across companies, megaprojects, science, technology, and visions.
Does it use AI?
The app includes live research through Valyu DeepResearch. The bundled atlas itself is curated and cached into structured data plus Markdown reports.
Can I self-host it?
Yes. You need Node.js, pnpm, a Mapbox public token, and a Valyu API key if you want live research.
Can I add more failures?
Yes. The repo includes scripts for checking curated entries, building draft entries, and merging them into the atlas.
Why study failed projects?
Because failed projects show constraints that success stories hide: financing, law, geography, safety, culture, trust, timing, and demand.
Where can I try it?
Try the hosted app or fork the code.
The thing I hope people do with it
I don't want Global Fail Map to make people less ambitious. In fact, I desire the opposite.
The kind that checks old ruins before pouring new concrete. That reads the trial record before believing the pitch and studies demand before building hardware. The kind that asks whether politics, land, financing, safety, and trust are part of the technical system.
If you are building something serious, use the map first. Maybe your idea is new or even older than you think.
Either way, the failures are waiting to teach you.
Top comments (0)