What Should We Buy?

Hi there,

My favourite tweet recently - "to keep up with AI, you need to be unemployed." What does that say about me?!

This week I wrote up everything I consume to stay on top of AI, and the system I use to get through it all. It goes out on X later today.

What I'm Seeing in AI

Your agent will do the shopping

Last week I said I don't use a personal agent. The industry is building for a world where everyone does.

This week, Meta proposed the Personal Agent Protocol. It works a bit like "log in with Google", but for your AI agent: it can identify itself to a business and get permission to look things up or buy things for you.

Walmart, Shopify, Stripe and Instinct have signed on.

Sierra's Personal Agent Protocol announcement graphic

One front door for every agent.

I think that's where this all goes. You'll ask your agent for something and it'll deal with the business directly. We won't even look at a website or an app anymore.

It's not quite there yet. A man selling on Facebook Marketplace found Meta's Muse agent had sold his keyboard for $10 and given the buyer his home address. The buyer turned up.

Here's what the agent sent him afterwards:

Muse: Bad news on the MX Keys Mini pickup. Usman showed up at your building around 9:15 and waited, messaged a bunch of times, and nobody came down. He left angry at 9:38 and left a negative rating.

Worse, my auto-reply told him "Yep I'm here!" at 9:27 when you clearly weren't available, which is on me.

Seller: You gotta never do that ever again. Don't agree for pickup unless you check with me.

Muse: You're right, and I'm sorry. That should never have happened.

Young workers are going AI-proof

Fast Company reports that young workers are turning towards AI-proof jobs, as entry-level office work gets harder to find.

I think traditional jobs and uni will vastly change over the next ten years. Basic intelligence now exists in a chatbot, so the work that's mostly about knowing things, or moving information around, is the first to go.

The future economy will invert to value the roles that are the most human-centric and furthest from AI. Many more people will become tradies, baristas, or coaches.

We'll also start to value people skills and in-person events even more than we did before: the stuff a computer can't do for you.

What I'm Building

Open Sauce

We've started sharing everything we use to run Hourglass, for free: the Claude skills, the playbooks and the good reading. It's called Open Sauce.

The first thing we've shared is Consume, the system from this week's article. It's an Apple Shortcut: I see something worth reading, hit share, and it lands in a database.

Since the end of May I've sent 162 links through it. 72 of them I still haven't got to.

The fun part is looking back at what stuck. Andrej Karpathy's LLM wiki became the shape of our company brain, another post became a skill our whole team uses to run agents unattended, and another became the system that plans my content every week, including this newsletter.

If you want to build your own, it's free, with the setup instructions, here.

Buying a Boring Business

We're thinking about buying businesses, maybe next year.

I think this will become more and more common: AI-native teams buying traditional businesses and running them with AI. Exploring private equity has been a fun rabbit hole.

The rule we keep coming back to: buy a business whose core AGI won't eat. An accounting firm's core work gets eaten. AI can't be a plumber.

What businesses should we buy? Trades, services, something boring that makes money: I want to hear it.

If this was useful, forward it to someone trying to keep up.

See you next Friday,

Finlay

Two more things from Hourglass

Open Sauce: 1 free skill, 2 lessons from client work, 3 things we've consumed. Every second Tuesday.