Everything I Consume to Build an AI-Native Company
文章目录
Staying on top of AI is a full-time job. There are YouTube videos, podcasts, essays, books, and about a thousand posts a day on Twitter.
But this is what's worked for me: the media & system that shaped how we start, run and keep up at Hourglass, our 7-figure AI-native company.
Everything in here, plus the system I use to keep up with it all, is free at thehourglass.ai/open-sauce.
The frontier is full of gaps
Paul Graham has a line in his essay How to Do Great Work that I think about a lot:
"Knowledge expands fractally, and from a distance its edges look smooth, but once you learn enough to get close to one, they turn out to be full of gaps."
AI is the biggest expansion of work we've ever had. It's the intelligence revolution. And once you get to the frontier, you can feel the gaps, hence why there's so much opportunity to make money right now.
Personal agents are the clearest example. Instinct, started by 23-year-old Noah Shinn, is a personal assistant you text or call. In September he raised $1 billion at a $10 billion valuation, four times its valuation a month earlier.
There are thousands of pieces of content from similar people who are currently finding different gaps and pushing the frontier in those places. Since AI allows hyper-personalisation, you don't need to copy other people's playbooks anymore. Now, you (or an agent) can consume it all, and apply the learnings to your own work.
Here are our favourite pieces, sorted by what they're for.
1. How to start a business
Paul Graham. After three failed products, Paul Graham's suggestion to become a consultant changed everything for me. It's his general thinking: talk to people, and promise to fix their work for just them. From Do Things that Don't Scale:
"Pick a single user and act as if they were consultants building something just for that one user... keep tweaking till you fit their needs perfectly, and you'll usually find you've made something other users want too."
That's how Hourglass started, and it's still how we work.
Tim Ferriss, The 4-Hour Workweek, also a massive one for me. The book is old, but the first half has insane value. He preaches that systemising a company to remove yourself as a bottleneck is essential to both the company's growth, and your own as a founder. Must read.
2. How to run an AI-native company
David Maister, Managing the Professional Service Firm (1993). We learn a lot from traditional consulting. Maister splits work into Brains (new problems), Grey Hair (problems you've seen before) and Procedure (problems you can run like a machine). Commoncog's breakdown is the best way in. It's now how we run Hourglass: we move every capability from Brains to Grey Hair to Procedure, and then into a product.
Sequoia, Services: The New Software (Julien Bek, March 2026). This one was the inspiration for what we do. "The next $1T company will be a software company masquerading as a services firm." My personal belief is that the age of the software product company is over. Soon, all software will have to 'learn' its user, rather than the other way round, and will have a layer of service whether that's via an agent or a human.
Every (Dan Shipper). Building an AI-native company and using it as the case study for the writing. It's the same playbook we're running with Hourglass in public, and a massive inspo for us. Their content is gold.
3. Staying on top of AI
Twitter. The best place for staying up to date with AI. Half of everything I save comes from X.
Situational Awareness** (Leopold Aschenbrenner) and AI 2027.** The two essays that kicked off modern AI doom/hype, predicting AGI would arrive in 2027. These essays made it so clear to me that the world hadn't priced in the AI diffusion gap, and how truly insane it's going to be.
There has never been a better time to start a company, get to the forefront of AI, and understand a technology that will change society as we know it. If you have any possibility to try it out, take a break from your career and learn it, you'll get massive rewards. That's what it's done for me.
The All-In Podcast. My favourite podcast, sometimes anti-doomer rants, but mostly a solid recap of the macro AI happenings from the week.
My First Million. A fun break from all the AI noise. They joke around, and they have a very empathetic view of how normal people will see AI. They're really good at making you think 'how does this fit into normal society?'.
4. Staying human
95% of the books I've read in the past five years are fiction. Vonnegut, Hemingway, Orwell, Dostoevsky, as well as some less classic stuff.
I read so much fiction, and I feel like classic fiction helps your mind unwind. You get to explore all these interesting ideas and typical human tropes, and understand our common failings.
Books like the Culture series by Iain M. Banks, Red Rising, or even just Vonnegut and 1984 take your brain to a creative place, to help solve problems and think about the future around AI, whether directly or not. The Culture does it directly: it's Elon's inspiration for how an AI utopia could look.
Our system for consuming content
The system that holds it all together: Consume.
It's an Apple Shortcut. I see something worth reading, I hit share, and it goes into a database. Once a week Claude goes through everything I've saved and decides what each thing is for: read it, save it to our company brain, share it with the team on Slack, review the tool, or turn it into something we build.
It's cool to look back and see how all these different articles became things we adopted:
-
Andrej Karpathy's LLM Wiki became the shape of our company brain, the thing we now build for clients.
-
Hamel Husain's "It's Hard to Eval" Is a Product Smell became our design rule for client builds: verification first.
-
Geoffrey Huntley's Ralph loops became a skill our whole team uses to run agents unattended.
-
Alex Lieberman's "I built a content machine" became the system that plans my content every week, including this article.
-
Peter Steinberger's test audit (he deleted around 400k lines of tests) is going into how we test our own code.
It's different for everyone
Overall, this is the media that helped me get to the bleeding edge of AI, start Hourglass and define how we run it.
We've started sharing everything we find, the skills, playbooks and reading, for free. It's called Open Sauce. Consume is in there too, with the setup instructions, so you can build your own: thehourglass.ai/open-sauce/consume. If you liked this content, subscribe to my newsletter.
At Hourglass we work with established businesses, from accounting firms to property and finance groups, auditing how their teams work and building the AI systems that do it. If that's you, book a call at thehourglass.ai, or DM me.
