Paying Attention #5: Fable Goes Dark, Agent-Native Tools, Moves on the Chessboard
Happy Sunday and especially happy Father’s Day to all the Dads out there!
This is my third Father’s day, and my first as a father of two. It’s the best possible thing you can do in your life. Yes, it’s extremely challenging. And at the same time, the most worthwhile challenge one can experience.
And now, for this week’s edition of “Paying Attention”…
1. Fable Goes Dark
What happened
The first full week of Claude Fable, arguably the most advanced form of technology ever released publicly, being unavailable for use. It’s become a real political football at this point. Coincidentally, the AI lab leaders and world political leaders were together for the G7 summit this week. President Trump was quoted as saying he “doesn’t view Anthropic as a national security threat”.
Anthropic executive Chris Ciauri said that Fable should be back “in the coming days”:
And today, the big news broke:
Why it held my attention
We are in uncharted territory. AI has gotten so powerful, so fast, that it is now being viewed as a national security concern. I thought this day would come. I did not think it would come this fast.
What I’m carrying forward
The precedents for how the wealthiest and most powerful democratic will regulate a solidifying now. The decisions that get made now will effect the future.
2. Agent-Native Tools
What happened
I discovered an open-source project called ‘Understand Anything’:
Why it held my attention
Tools like this allow you to do things that would have been impossible before. Or, at least very difficult and time consuming. Now you can instantly gain an understanding of vast amounts of information.
What I’m carrying forward
I used ‘Understand Anything’ all night last night after discovering it. Applying tools like this to your real work can already provide material value to your life.
3. Moves on the Chessboard
What happened
GLM-5.2 was released and it’s the best open-source model available according to many. I haven’t used. Also, the CEO of Z.ai, the company that develops the GLM models, said to Elon Musk that China will have Mythos level models before 2027.
There were also some big personnel moves this week. Noam Shazeer, co-author on the original Transformers paper “Attention is All You Need” left Google DeepMind for OpenAI. Shazeer’s company Character.ai was acquired by Google for $2.7B to bring Noam back into the company.
John Jumper, Nobel prize winner for his contributions to AlphaFold, also left Google DeepMind for Anthropic.
Why it held my attention
These events are a hint to the answer of two very important questions:
How far is open-source from the proprietary frontier?
Are OpenAI and Anthropic the only viable contenders for AGI?
What I’m carrying forward
AI is starting to get serious. This week showed us how governments are starting to take it seriously. It’s only going to get more serious from here.