Hi there,
You’re receiving this because you signed up for the Hourglass Digital newsletter. I haven’t sent one out in the past few months because a lot has changed this year.
We became a proper company, renamed to Hourglass AI, and I brought on a Co-Founder + two employees. We’ve now worked with 35 companies, and many more to come 🤞.
Why Personal?
A company newsletter felt salesy, and didn’t fit the purpose I wanted. The aim is simply to share the trends I’m seeing in AI from using it 6+ hours a day and helping businesses use it.
Expect something a lot more personal, and not necessarily 100% about AI.
We also have a company-specific newsletter, that you can subscribe to here to see our wins, losses, and revenue, mainly to publicly keep us accountable. Also, monthly seemed too far apart to keep you up to date with how fast AI is moving: expect me in your inbox every Friday instead :).
What I’m Seeing in AI
In the past little while, the AI buzz has gone a little quiet. I think it’s making people complacent.
It’s easy to argue that less news and slow progress means AI isn’t working out, and all of our jobs are safe, but the truth is, we don’t know that. We could be one day from a game-changing development, so I urge you all to keep trying different things.
That being said, it does feel like the models are plateauing a little. This week, Anthropic released Claude Opus 4.8, their newest, most powerful AI model. OpenAI’s best model currently is GPT-5.5. 4.8 feels like at most an incremental jump.
It seems that recently, new model releases are optimisations, and not the massive jumps we’ve had for the past four years (can’t believe it’s been that long). In the context of work, it seems like the real improvements aren’t the AI getting better themselves, but all the tools and code around it.
84% of Australians use AI every day but only about 7% are using it well.
Whether the exact number holds or not, it matches what I see. The gap between people using AI and people building with it is widening fast, and this is the important statistic, not the model benchmark scores.
What I’m Building
I also open-sourced some of our company AI skills on X, which went semi-viral.
My post - 4 times more saves than likes 🤨
This showed me that the hardest part about AI for all of us is finding out what’s out there, figuring out what it can do, and learning the best way to use it in your life.
People want the perfect skill or outcome without the learning, and I honestly don’t blame them. I think this is the right approach, and you’re better off spending the time learning from someone and applying the AI skills to make you even better at what you’re already good at (i.e. your job, study, hobby).
This shaped our company strategy and direction a little; we’re exploring how we can lean more into a community/product setup that automatically keeps people and businesses on the best AI practices that we’ve found from staying on the bleeding edge, reading, and trying everything.
That’s it for this week.
I’d genuinely love to hear what you want more of, or less of. Any questions or arguments are very welcome: just hit reply.
And if someone in your world would get something out of this, send it their way.
See you next Friday,
Finlay
Explore the rest of the newsletters at finlayekins.com/writing.
P.S. If you think of a better name than ‘Finlay’s Newsletter’, let me know.