An atlas of the institutions that forged the modern world

Innovation Engines

Thirteen places — labs, universities, accelerators, mafias, agencies — that produced an outsized share of everything you use today.

Begin the tourSee the patterns
13
Institutions
80+
World-changing inventions
100Y
Years of accumulated DNA
$30T+
Value created downstream
Scroll
Manifesto

Why these thirteen

Most of what we call 'technology' came out of a remarkably small number of places. A handful of corporate research labs that ran on monopoly rents. A handful of universities geographically lucky enough to sit next to capital. A few accelerators, a few mafias, one or two government agencies willing to spend other people's money on long shots. Read carefully, the lineage chart of modern computing is shorter than you think — and the pattern repeats. This site is a study of that pattern.

The thirteen

Profiles in depth

Founders, inventions, spinoffs, culture, the quote that sums it up.

Lineage

The genealogy of modern tech

Where one institution ends, the next begins. Walk the tree.

Bell LabsShockley Semi.FairchildIntelAMDKleiner PerkinsNvidiaDARPAInternetGPSXerox PARCApple MacMS WindowsAdobe3Com (Ethernet)StanfordHPGoogleCiscoSun Micro.MITRSAAkamaiCMUUC BerkeleyBSD → macOS, iOSPostgresDECWindows NTAltaVistaPayPalTeslaSpaceXLinkedInYouTubePalantirFounders FundY CombinatorStripeAirbnbCoinbaseOpenAIAnthropicxAIa16z

Solid arrows: direct founding lineage. Dashed: significant influence or alumni-led contribution. Hover any node to isolate its network.

What they share

Anatomy of an innovation engine

Six traits that show up everywhere. Three myths that don't.

01

Concentrated, voluntary talent

Every engine has a population of obsessed people choosing to be there at below-market terms in exchange for proximity to peers and problems.

Bell Labs hallways · YC Demo Day dinners
02

Long time horizons

Most great inventions sit in a lab for 5–20 years before they reach a customer. Engines that quarterly-report die.

Transistor: 1947→1958 commercial · GPS: 1973→2000 civilian
03

Funder with patient capital

Bell had AT&T monopoly. PARC had Xerox's cash cow. DARPA has the federal budget. YC has its own balance sheet. Without a sheltered budget, the engine wobbles.

AT&T regulated profit · DoD appropriations · Endowments
04

Open culture, defended

Free flow of ideas internally. Open publishing externally. But protected from the kind of bureaucracy that would discipline either.

Bell Labs papers · PARC open visit days · arXiv-era OpenAI
05

Spinoff DNA

An engine is judged less by its own products than by the companies its alumni start within five years of leaving.

Fairchild → 100+ chip cos · PayPal Mafia → $2T+
06

Mythology

A self-narrating engine recruits better. 'The PayPal Mafia.' 'Bell Labs.' 'PARC.' The story is part of the asset.

Toga photoshoot · Idea Factory · Dealers of Lightning
Three myths that don't hold up

Myth: Lone geniuses

Almost none of these breakthroughs were one person. Shannon had Bell. Page had Brin. Sutskever had OpenAI. Pairs and rooms, not loners.

Myth: Garages built this

The garage is downstream. Upstream is a university lab, a corporate research budget, a DARPA grant.

Myth: VC is the engine

VC is the gearbox, not the combustion. It rides on top of decades of pre-funded research it didn't pay for.

What comes next

Where the next engines are forming

AI labs. Open-source DAOs. New geographies. Bet at your own risk.

01

Frontier AI labs

OpenAI, Anthropic, Google DeepMind, xAI

The new Bell Labs — but with their work shipping in public weekly. Compute-intensive, talent-concentrated, mission-marketed.

02

Open-source protocols

Ethereum, Bitcoin, NixOS, Linux Foundation

What BSD did for OSes, public protocols may do for finance and coordination. Genuine engines, just slower to credit.

03

Government-backed mega-labs

ARIA (UK), SPRIND (Germany), China's CAS labs

Trying to recreate DARPA as a non-American export. Outcomes pending — DARPA copy-paste is harder than it looks.

04

Distributed talent rooms

Hugging Face, Replicate, Cursor's user base, vibe-coded hubs

Internet-native engines: no campus, but very high collision rate. Closer to Stanford-as-soil than to PARC-as-fortress.

How to use this

For founders, researchers, policymakers

Five concrete lessons, each backed by a precedent above.

For founders
01 /

Pick your room more carefully than your idea. The room compounds; the idea probably changes.

02 /

Optimize for being in someone's later genealogy chart. The PayPal Mafia photo is only valuable in hindsight.

For researchers
01 /

Publish like Bell Labs did: clearly, completely, with the failures. Reputation flywheels on legible work.

02 /

Choose problems older than your tenure clock. Five-year horizons are too short for most real breakthroughs.

For policymakers
01 /

DARPA-shaped funding is the most asymmetric tool a government has. Program-manager autonomy is the secret, not the budget.

02 /

Universities only become engines when they're next to capital and allowed to be promiscuous about it. The geography is policy.

03 /

Monopoly profits funded Bell, AT&T, and IBM research. The trade-off is real — competition policy decides whether the next Bell Labs is possible.