#006 - Are you ready for AGI, Winnetou?
It was 1885. London was a shitty place. Literally.
Horses ran the city. Within a decade, the capital city relied on roughly 300,000 of them to move people and goods. Then things started to change.
Benz filed his patent in 1886. Daimler and Maybach fitted engines to almost everything on wheels. Peugeot rolled a steam trike into the 1889 Paris Expo, then switched to Daimler engines the following year. Across the Atlantic, Olds' Curved Dash became the first mass-market car in 1901. And because humans are hilarious, an inventor even proposed the "Horsey Horseless"—a vehicle with a wooden horse head on the front so it wouldn't scare… horses. You can't make this up.
For years, the two systems (horses and cars) ran in parallel. Cities still used horse-drawn wagons to deliver parts and people to early car plants and dealers. Carriage giants like Durant-Dort supplied the upstart auto firms, then morphed into them. Model Ts left the Piquette Avenue Plant by rail. Yet the streets around the factories and showrooms were still crowded with horses. Horse muscle moved metal, until it didn't.
And then it stopped. Sometimes in a glide. Sometimes, like a cliff.
The "mechanical horse" erased the working horse. Today, London has ~200 horses in the city—mainly ceremonial and mounted police—serving around 9.8 million people. In 1900, the city had ~300,000 horses for a population of around 6.5 million. That's on the order of 2,000× fewer horses per capita than at the peak.
Read that again. Two thousand times.
History is repeating itself now.
We work beside a new class of machines that will replace tasks, teams, and entire tiers of jobs. Not in theory. In payroll.
For the first time, a tool touches thinking—analysis, planning, even basic empathy. Steam replaced the horse. The internet replaced paper. Software is now eating cognition. You are next unless you move.
What's Coming (And What To Do About It)
Let's keep this clean.
Truth one: when technologies collide, sentiment loses to physics. Don't calm yourself with the "people will always…" stories. History is a museum of retired certainties.
Truth two: your mindset is the first moat. The timing is unclear. Sam Altman says jobs will change fast—some "totally go away," while many get supercharged. Demis Hassabis is a cautious optimist, calling the wave "bigger and maybe faster" than the Industrial Revolution, yet betting on human adaptability. Yuval Harari points to the hard part we avoid naming: identity shock and the need to unlearn and relearn, again and again. For executives: Strengthen your resilience first. Your CV comes second.
Truth three: become hard to replace. Not in a slogan. In the market. Pick a lane where you can be top-100 in the world or top-5 in your country. Narrow until it's true. Get in touch if you need a mentor or coach.
A Note On Pace And Pain
Executives love smooth curves. You won't get one. Adoption looks like a step function. A client standardizes on AI agents. A regulator green-lights a class of models. A competitor enters the market with costs reduced by 70%. One quarter, you're fine. Next quarter, you're a case study.
Yes, there will be new jobs. Join the “AI-native” generation that treats these tools like oxygen. And many jobs will go away. The mix shifts. Your goal is to keep many paths open as work evolves.
The History Lesson You Don't Want
Did horses survive the steam age?
A small, curated sample did—in stables, shows, and parades.
Did entire cultures survive technology shocks? Yes—many did, but often only in small enclaves under the winners' rules.
I'm not fear-mongering. I'm reminding you that technology reallocates dignity. If you don't move, the market will move you. The horses didn't "reskill." The coachmakers who became automakers did. The carriage parts suppliers who sold into Buick did. The rest became history.
If You're Running A Company
Make AI a standing item on the exec agenda. Fund three live pilots tied to P&L this quarter. Put a builder and a skeptic in the same room. Change compensation to reward shipped AI use, not slideware. Publish a one-page AI Constitution so your people know what's sacred and what's fair game.
Personally, set a 90-minute weekly "AI block" in your calendar. One hour practicing AI on your core workflows, thirty minutes reading two papers or release notes. I'll tell you how. That's it. Repeat. Small, relentless steps beat big, performative ones.
Remember the horses' math. Obsolescence scales quietly, then all at once. Don't be the last horse standing.
You wanted clarity. Here it is.
Act like a species that adapts, because you are.
Your move.
For now, read my newest blog post:
👉 How Market Leaders Outpace the Competition with Agentic AI.
