Heya and welcome back to Five Things AI!
Yesterday I gave a quick talk about Agentic AI and what happens after the pilot - as the pilot always looks awesome and then we need to figure out how to run agents in production. Who is responsible for the whole process? What do we do if it breaks? How do we make sure the Agents always have up-to-date documentation? For me, the most interesting part of the talk I gave was this: people asked my about me giving a talk at a competitor’s event. I really didn’t see it that way. Sure, Octonomy.ai and my Agentic AI platform tectic.ai are both Agentic AI platforms that share the same underlying concepts, but yet we are so different in what we do and how we do it. The market is growing so quickly that I think the differentiation will be different in the future: pure-play Agentic or bolted-on Agentic? Pure-play obviously will be better and both Octonomy and tectic are betting on the fact that this sector will grow tremendously in the near future.
How to Regulate the Ghosts
The frontier of AI research needs the most oversight, but only a handful of companies are operating in that space. That makes regulation easier. AI companies pursuing lower-risk capabilities—a chatbot that recommends restaurants, an image-recognition model that improves municipal recycling —don’t need the same oversight as OpenAI. The EU Artificial Intelligence Act already sorts systems this way: Cross a set amount of computing power while training your model, and you’re automatically in the strictest tier of oversight—no argument needed.
None of these actions tell a lab what it can or can’t build or how fast it can build it. They regulate visibility and accountability around safety and deployment while leaving lots of room for future debate.
I actually think that the unwillingness of the US government to regulate AI in any meaningful way is hurting the adoption rate of AI. The EU AI Act is by no means perfect, but it allows for regulatory sandboxes that make a ton of sense as they provide guided exemptions for fast moving AI companies.
How Accurate Have AI Progress Forecasts Been So Far?
In brief, across our studies, forecasters dramatically underestimate AI progress on benchmarks, have a more mixed track record on AI adoption and diffusion though often lean toward underestimation, and there is not yet enough evidence to assess their track record on predicting macro-scale economic and societal impacts of AI.
The old saying that “prediction is very difficult, especially about the future” is still pretty much true when it comes to AI development.
Analyzing Jev, a new AI model
Behind the impressive marketing is a very interesting model. Jev is a close relative of LLMs like GPT, Claude, or DeepSeek, but is actually more similar to the BERT series of models, which are used for classification, not generation. (If you've ever wondered why someone calls out a difference between “AI” and “GenAI”, that's the distinction they mean).
I think Jev opened up the doors to a whole different way of thinking about how we can use AI. It’s been an impressive launch and so many people are building interesting things around it.
These Tiny Startups Are Getting Even Smaller With Help From AI
Plenty of startups increase head count at a rapid clip as they aim for billion-dollar valuations and fight for top engineering talent. But a subset of early-stage founders is embracing an ethos derived from an extreme reliance on artificial intelligence: Stay small for as long as possible.
Before hiring, these founders are asking if AI can do the job instead. Some have laid off employees or cut ties with contractors and agencies because of AI. Time will tell if rising sales—or outside investors—force these startups to increase human head counts too.
This will be the new normal for startups - spending money on tokens is easier than hiring and paying humans.
Is Artificial Intelligence Commoditized?
Everyone is dizzy with excitement about Artificial Intelligence. On the way in, everyone wants to invest. CAPEX spending is off the charts.
But no one is asking what this looks like on the other side from an economic stand point.
Will there be sufficient payback to justify the expense?
Will investors see an acceptable return?
Will AI become a commoditized service?
The introduction of any new General Purpose Technology (GPT) will always make a huge societal impact. But that is rarely enough to support an investment thesis.
As the open weight models are catching up to the frontier models, I see less and less necessity for paying a premium price. This will be good for adoption of AI, but not necessarily for the frontier labs.
If you missed last week’s edition of Five Things AI, you can read it here:
That’s it for Five Things AI this week! 🤖
— Nico