While VCs chase the next LLM wrapper, one founder is printing money with HTML and a passion project about maps from 1823.
The Summary
- Craig Campbell, ex-Meta engineer, turned down blank-check AI funding to build Past Maps, a website selling access to historical maps
- He walked away in 2022, right as the AI boom peaked and VCs were desperate to fund anything with "AI" in the pitch deck
- The contrarian bet on "old school web" is working, running counter to the narrative that Google Zero kills organic web traffic
The Signal
Craig Campbell had the golden ticket. Fresh off selling his Shopify e-commerce tool, sitting across from VCs offering blank checks, watching everyone around him pivot to AI in 2022. He said no. Instead, he built Past Maps, a website that does one thing: lets you browse and buy access to scans of historical maps.
Not an AI-powered map generator. Not a blockchain-based cartography DAO. A website. With maps. From before your great-grandparents were born.
"I had my prior VC investors breathing down my neck, going 'start something else. We'll write you a blank check.'"
The timing matters. 2022 was ChatGPT's debut year. Every founder with a pulse was either building an AI company or pretending their existing company was secretly an AI company all along. Campbell watched the feeding frenzy and went the other direction. He bet on search traffic, organic discovery, and people actually wanting to pay for digital things that aren't subscriptions to yet another productivity tool.
The Google Zero thesis says this shouldn't work. When AI answers appear above search results, when ChatGPT summarizes instead of linking, when traffic to websites drops because people never leave the search page, you're supposed to be dead. Campbell's site isn't dead. The piece says he's "grown" it, past tense, implying success without stating exact numbers.
What makes this work:
- Specificity wins. Historical maps are too niche for AI to train comprehensively on, too visual for text summaries to replace
- Direct value exchange. You want this thing, you pay for this thing. No ad model depending on eyeballs
- The anti-SaaS model. One purchase, you own it. No monthly retention anxiety
This is the opposite of how you're supposed to build in 2024-2026. No AI moat. No network effects. No recurring revenue. No venture scale. Just a thing people want, sold directly, built by one person who knows how to code.
The Implication
The agent economy everyone's building toward assumes AI will intermediate everything. Campbell's bet is that there's still huge value in the unintermediated web. Things people seek out directly. Things too specific or visual for summarization. Things worth owning outright.
Watch for more founders walking away from AI money to build these kinds of focused, profitable websites. Not because they're anti-AI, but because they see the opportunity in what AI can't or won't do well. The future might not be AI doing everything. It might be AI doing the general stuff while humans build weird, specific, profitable corners of the internet that nobody thought to automate.