Hi there,
I think recently, it’s become clear that AI just works now. And the people who have felt that ‘aha’ moment are massively outpacing those who haven’t.
However, now that it works, the bigger problem is how do we actually use it well, and what do we do with it?
What I’m Seeing in AI
The Show is About to Begin
AI just crossed the line from demo to doing real work, and in some cases doing it better than us on the first go.
Stanford’s latest report had the stat that agrees: AI agents went from 12% to 66% success on real computer tasks in the last year. Stanford 2026 AI Index.
AI nails two thirds of tasks and is closing in on humans
That matches what I keep seeing in person. People are genuinely surprised at how much it can take off their plate, once they get over the hurdle of setting it up.
So if you’re still using AI in a browser tab, this is the week to download the Claude desktop app or OpenAI’s Codex app and actually let it loose: sorting your emails, writing your documents, building you a website.
It feels like we’re back at day zero. The thing works now, so the only question left is what we do with it.
Human Agency
For me, the bigger shift isn’t the models, it’s what expert-level AI does to us.
I was listening to the All-In Podcast last week (on AI and economy) and the hosts discussed something I really believe - no human who ever lived has come close to their true capacity. Human agency is basically limitless, and the right tools and mindset are how you reach more of it.
AI is the biggest unlock of human agency in history. If the thing capping people was always the skills, time, and know-how, then everyone now has all of those things. It makes me genuinely excited for all the people who can now take a newer, braver, more interesting path than the one they thought they were stuck on.
There’s no ceiling
What I’m Building
Rebuilding Glassy
We’re rebuilding our own AI agent from the ground up. We have an assistant called Glassy that runs a lot of our day to day, and the off-the-shelf pieces underneath it (OpenClaw, Hermes) got too annoying to bend to what we actually needed. So we’re going fully custom.
It’s the same custom-software-wins point I keep mentioning - tools now learn how you work, and not the other way round.
If We Were All like Elon (but normal)
I’m currently reading the Elon Musk biography. Whatever you make of him, it’s a study in someone operating at the very edge of what one person will attempt. He has almost no regard for anything except having a generational impact on our species.
It makes you realise that if everyone carried even a fraction of that delusion and self-belief, the world could be unrecognisable in ten years.
To tie it all together, what would you do with your life if you could genuinely do anything? Intelligence and skills aren’t a limit. Now the only limit is what we can imagine, and what we’re brave enough to try.
Send this to someone who needs to think bigger.
See you next Friday (and cmon socceroos),
Finlay