Hi there,
This week was the Hourglass team offsite. I’ll throw in what we did, what we're aiming for by the end of the year, and what we think happens next with AI.
First, two AI stories from the week
What I'm Seeing in AI
The most valuable room in the world
The device behind the TV
An OpenAI engineer allegedly hid a listening device in an executive conference room at their San Francisco office. It was found on 30 September, wrapped in toilet paper and plugged into the USB port on the back of a TV.
The device is a hobbyist board with a microphone and a camera. OpenAI says it used its own AI models to trace the device back to him. He was fired and arrested on 9 October by the FBI.
I think it highlights to me that the conversations inside big AI labs right now might be some of the most important and valuable of any in the world.
AI is now building AI
The State of AI Report came out last week, stating in August, Claude led 26% of Anthropic's own model research. In February it was under 1%.
This is recursive self-improvement (RSI): AI helping build the next AI, which helps build the one after that, and so on. All of the big AI companies are racing to achieve this, because it will mean super-intelligent AI.
OpenAI's stated goal is an automated research intern, then a fully automated research workforce. Anthropic co-founder Jack Clark puts the chance of a model fully training its own successor by the end of 2028 at over 60%.
At our offsite we voted 16/25 that every lab gets there by 2028. I think we could have it by the end of the year.
What I'm Building
The offsite
Our company offsite was a mix of strategy and fun.
The aim was to disagree hard, then commit to a path. We started with all of our predictions for the future of work and AI (see below), and then narrowed it down to our company and finishing off the year.
Our goals for the end of the year:
One North Star: 10 healthy retainers. Clients who stay with us month to month and actually use what we build
Three pillars, each with an owner: winning clients (Batko + Matt), building for them (Jeremy and Anna), and keeping them (Finlay)
One number each. A metric that you can define your work on every day
We plan on taking fewer, but bigger and bolder bets to become the biggest AI name in Australia 🇦🇺
What we think happens next
45 of our predictions on the future
We wrote up any and every prediction we had for the future and grouped them into categories. We then each voted on them out of 5.
Our top score, 23 out of 25: personal brands become the distribution method.
People pick companies by the people behind them, so every one of us needs one, not just the founders. We predict in the next 2 years, you will know the name and face of most successful CEOs, even at massive companies.
A few others about AI in everyday life:
Your mum texts an agent to buy something
A humanoid robot in your home: our median guess was six years
A human-only social network takes half the market: 9/25. Most of us think the feeds will just filter AI out instead
Will your mum be buying toilet paper through an agent in three years? What do you think?
And if you know someone who'd enjoy this, forward it on.
See you next Friday,
Finlay
Explore the rest of my writing here → finlayekins.com/writing
Join our monthly company newsletter here → thehourglass.ai/about
Want this running in your company? → thehourglass.ai/call/finlay