Conference reviews
The main talks, ideas and limitations — reviewed from the complete programme
Journal topics
AI sped up vulnerability discovery. Who can fix the flaws?
At Black Hat USA 2026, Yan Shoshitaishvili described more than a thousand findings in the Linux kernel. His experiments show how task definition changes results, and why functioning code still needs security validation.
HubSpot lost 80% of its traffic. Why did leads grow?
At UNBOUND 2026, Kipp Bodnar and Kieran Flanagan explained how they are changing marketing as buyers turn to AI answers. Their approach centres on distinctive content and connecting advertising to sales outcomes.
WWDC 2026: what Apple showed about local AI and agents
Across six recordings, Apple describes an agent not as one assistant, but as an application with a model, tools, history, tests and limits on access.
IJCAI–ECAI 2026: what the main stage said
The Bremen conference was not a contest over one winning model. Its full main-stage recordings returned to rules between agents, data with different human answers, the physics of robots, and who can take part in research at all.
ModCon 2026: why one good AI model is no longer enough
Qualcomm paid nearly $4 billion not for another model, but for a way to run many models across many chips. Modular’s conference showed why that still depends on memory, networking, electricity and the people holding the system together.