After Web 2.0

1. From ARPANET to the Wild West

The class opens with a sweep of internet history: ARPANET in the military and academia, walled gardens like CompuServe (1979) and AOL, then the Mosaic browser in 1993 and Netscape's founding in 1994 blowing the web open. What people did online kept shifting, the user base grew roughly 20x after the late '90s, and mobile was widely seen as the next era.

• Early '90s: file transfer and email for academics — you practically needed a CS degree just to get online
• Late '90s: web browsing and peer-to-peer take over, with only about 50 million people online
• By 2010: video passes half of all traffic, billions are connected, and mobile looms as the next platform

Thiel's frame: for two decades the internet has been a Wild West. The world of atoms is heavily regulated — finance was arguably 'too innovative' and got clamped down — while the world of bits mostly is not, which is why computing stayed fertile. The glaring exception is patents, which function like regulation aimed squarely at the small: no regulator can shut you down for being tiny, but a patent holder can. The SOPA and PIPA fights showed the mounting pressure to tame the frontier, because the Wild West makes people uncomfortable.

2. When Will the Future Arrive?

History is littered with confident predictions that missed: Lord Kelvin declared heavier-than-air flight impossible in 1895; the U.S. patent commissioner supposedly announced in 1899 that everything had already been invented. The gaps run the other way too — the Chinese built rockets in the 13th century, yet the moon landing took another seven centuries; the Apple Newton shipped in 1993, fifteen years before the iPhone made the idea work. Napster arrived too early and too confrontational; Spotify nailed the timing.

All the ideas of the '90s were basically correct. They were just too early.

• Being early burns capital while you wait for the market to show up
• Your architecture is built for a world that doesn't exist yet — and is outdated by the time it does
• The long wait corrodes company culture
• Founders rarely get a second shot: whoever built Friendster didn't get to build Facebook

The 2000 crash split the industry into two populations. Survivors carry burned-on-the-stove scarring and hunt for bubbles everywhere, like the 1929 generation that never trusted stocks again — scarring that only dies off, never fades. Those who were too young escaped it entirely: Mark Zuckerberg, in junior high during the bust, reportedly had to ask what Netscape ever did. Thiel's timing heuristic is a surfing metaphor — you cannot wait until the wave is confirmed, and it is better to occasionally paddle for a wave that never comes than to miss the big one.

You have to paddle early, and then let the wave catch you.

3. Software Is Eating the World — Three Versions

Thiel adds a 2x2: do you compete with computers or work with them, and do you compete with globalization (China) or work with it? The winning quadrant works with both. Near-term targets for software: healthcare records and analytics, computerized education, labor marketplaces like Uber and TaskRabbit that route around regulated models, and eventually law. In practice, a16z invests in no cleantech and no biotech — only companies that would collapse if you removed the software team — and expects incumbents to fight the transition every step of the way.

The clearest near-term wave is e-commerce 2.0. Version 1.0 was search-driven — Amazon and eBay; 2.0 companies like Warby Parker and Airbnb win on deep understanding of consumer behavior. With roughly 2.5 billion people online versus 50 million in the '90s, formerly absurd ideas now work: Diapers.com sold to Amazon for $450 million, Golfballs.com thrives, and Webvan-style grocery delivery is returning city by city. Retail's high fixed costs and razor-thin margins make it fragile — Andreessen jokes his 'fake hedge fund' is short retail, long e-commerce.

4. Disrupt Without Getting Crushed

Disruptive kids get sent to the principal's office: Napster won the users but got crushed by the industry it attacked, and the record labels celebrated — prematurely. Spotify took the opposite approach: it writes massive checks to the labels, launched first in low-CD markets like Sweden, and staggers its contract expirations so the labels cannot coordinate a rate hike — the trap that squeezed Netflix when content providers jacked up prices in unison. The risk remains that incumbents take the money, demand equity, and still try to eliminate you.

1. 1

   Open with money

   A conversation that starts with 'here's a check' goes better than one that starts with 'we're replacing you.' Revenue softens the disruption blow.

2. 2

   Enter indirectly

   Launch where you threaten incumbents least — Spotify started in geographies where CD sales were already weak, proving the model before facing the main fortress.

3. 3

   Stagger your contracts

   Rolling expiration dates stop suppliers from coordinating a simultaneous squeeze; incumbents counter with shorter deals and non-dilutable equity stakes.

New gatekeepers are forming even inside tech: on mobile, great distribution tricks get banned and then copied by Apple and Android, and Apple blocked iOS apps from integrating Dropbox — even a large startup can be stopped dead. As for patents, the system is broken — examiners can no longer judge novelty, and big companies stockpile thousands of patents a year — but Thiel reframes the pain: patent trouble means you built something worth suing over. It's a problem you want to have.

5. Distribution, Boards, and the Learnable CEO

Andreessen's number one reason for rejecting startups: a great product with no distribution strategy, usually disguised as a 'viral marketing strategy.' Founders cite Salesforce as proof that products sell themselves — but Salesforce runs a huge sales force; its tagline is 'No software,' not 'No sales.' Well-run software companies pair a great engineering culture with a great sales culture, with one law strictly enforced: sales never orders the product team around, or you decay into a consulting shop.

• Thiel on boards: three people is optimal — the bigger the board, the weaker the oversight; a 50-person nonprofit board means the manager answers to no one
• Andreessen has never seen a contentious board vote — what kills companies is people, not process; startups are sausage factories, and founders under-vet their VCs
• When things go wrong, boards bias toward 'doing something' — which is often worse than the problem itself

a16z's most controversial position: CEO is a learnable skill, not something shrink-wrapped from a 'world-class CEO' mill. The counter-evidence to conventional VC wisdom is decisive — Microsoft, Google, and Facebook were built by inexperienced product founders who learned on the job. The curriculum: learn to manage managers (which scales, unlike managing people), enough law to stay out of jail, enough finance to raise money, enough sales to sell. Sales-guy CEOs can optimize a company for two to four years before it hollows out; you cannot swap a Pepsi marketing executive in for Steve Jobs — and Apple's real edge is its software, not its hardware.

On where students should start: if outcomes follow a power law, picking the single best company matters far more than choosing between 'founder vs. employee.' Big-company alumni rarely start startups — everything works automatically there, so you never see the machinery. Andreessen's proof from his 1991 IBM internship, 14 levels below the CEO in a 400,000-person org, is the sharpest line of the night.

The skill that you learn at IBM is how to exist at IBM. It's completely self-referential.

Then vs. now (2026)

Self-check quiz

Pick an answer to reveal the explanation.