I’ve spent most of the last month working and traveling across the U.S.
It’s been brilliant. It’s also been busy.
So I haven’t been able to shoot a new demo – or even post a lot about AI.
That changes next week.
But today I want to give you 4 things that are happening in AI right now that should make you stop mid-scroll and say:
what the ?
1. AI is starting to help build itself.
Anthropic says Claude helped implement a complex Claude Code feature, with about 70% of the code written autonomously by Claude. Let that sink in. The tools are no longer just being used. They’re helping build the next version of themselves.
2. AI is not just smart anymore. It’s starting to look creative and empathetic too.
A JAMA Internal Medicine study found chatbot responses to patient questions were rated higher than physicians’ responses for both quality and empathy. Separate research in Scientific Reports found AI chatbots outperformed the average human on a standard creativity task, even though the very best humans still came out on top.
3. Most organisations are still using AI at the kiddie-pool end.
Summaries. Meeting notes. First drafts.
Useful? Yes.
Transformational? Not even close.
Meanwhile, frontier models are writing code, reasoning across complex tasks, and performing credibly in areas we used to think were deeply human. The gap between current capability and current use is still enormous.
I’ve read that some estimates (sorry I don’t have a source) say that we are only using about 1% of AI’s current capability. This doesn’t address the capability that will be developed. That’s what we have available to all of us commercially right now
That may be the biggest story of all.
4. Some models are now being held back because of what they could enable.
Anthropic says Claude Mythos Preview is especially capable at computer security tasks, and instead of a broad release, it launched Project Glasswing to give limited partners access to help secure critical software. We’ve officially entered the era where “too powerful to release widely” is not sci-fi language. It’s an actual product decision.
That’s where we are now:
- AI is helping build AI.
- It is showing signs of empathy and creativity.
- Most companies are still barely scratching the surface.
- And some of the most capable systems are being restricted before full release.
This space is moving faster than most leaders realise.
More next week.
