GLM-5.3 is the highest-scoring open-weight model on the Cyber ECI with a score of 160, surpassing GPT-5.4 released in March. The post gives a 6–7 month gap between closed frontier and open-weight cyber capabilities.
Key points
Score: GLM-5.3 achieved 160 on the Cyber ECI.
Release: GPT-5.4 was released in March.
Sources
Frida new iOS personal agent launches on TestFlight
Frida is a new iOS personal agent that lets users schedule, book, shop, and travel from a minimal interface that requires explicit approval. It is currently available on TestFlight to a small group of early users.
Key points
Scheduling, booking, shopping, travel appear in an intuitive, minimal interface you can actually act on, and nothing happens without your approval.
Currently on TestFlight with a small group of early users.
