In partnership with
News of the day
1. AI leaders debate intelligence: Hassabis sees singularity approaching, LeCun says current AI isn't intelligent, Vinyals notes limitations in learning and innovation. → Read more
2. Anthropic's acquisition of Stainless, a key SDK generator, removes a shared supplier for rivals like OpenAI and Google, shifting the developer toolchain landscape. → Read more
3. Agentic AI accelerates development but strains data governance. New strategies using MCP and synthetic data are crucial for real-time compliance and trusted AI. → Read more
4. Anthropic likely to continue supplying Claude AI to NSA despite Pentagon's supply chain risk flag due to chip shortages. → Read more
Our take
Hi Dotikers!
While Yann LeCun was once again explaining that LLMs aren't really intelligent, an internal OpenAI model was demolishing an 80-year-old Erdős conjecture.
Welcome to the week the AGI debate completely went off the rails.
Three of the field's biggest voices spoke up within days of each other. Three readings of the present. Three futures that can't all be true.
Demis Hassabis, closing the Google I/O 2026 keynote, drops a word lab CEOs never use this way in public: "singularity." "Humanity is standing in the foothills of the singularity." For a man known for his almost pathological caution, this is a historic shift in tone. He puts AGI within five years, and calls it "ten industrial revolutions compressed into a single decade."
LeCun, for his part, dusts off Piaget: "Intelligence is not what you know, it's what you do when you don't know." Verdict: LLMs accumulate, recombine, regurgitate. But put them in front of anything genuinely new and they collapse. He's been working for years on something else: systems that would learn the way a child learns by watching a ball fall, not by swallowing three million papers.
Except that five days earlier, OpenAI announced exactly what he says is impossible. The unit distance problem in the plane, posed by Erdős in 1946: for 80 years, everyone believed square grids were the best possible construction. An internal model found an entirely new family that does better, and along the way revealed an unexpected connection between number theory and discrete geometry. First time an AI has autonomously solved an open problem central to a field of mathematics.
That leaves Oriol Vinyals (co-lead of Gemini), who holds the most honest position: seven years ago, he says, he would have called these models AGI without hesitation. Today, he points to the two missing pieces: learning continuously from experience, and producing real discoveries. On that second one, OpenAI just answered him in real time.
This isn't a technical disagreement. It's the sign that we're living through a moment when the map no longer matches the territory, and even the people drawing the best maps are saying it out loud.
Alex.
The IT strategy every team needs for 2026
2026 will redefine IT as a strategic driver of global growth. Automation, AI-driven support, unified platforms, and zero-trust security are becoming standard, especially for distributed teams. This toolkit helps IT and HR leaders assess readiness, define goals, and build a scalable, audit-ready IT strategy for the year ahead. Learn what’s changing and how to prepare.
Meme of the day
