GLM-5.3: frontier coding with emergent cyber capabilities
Z.ai's GLM-5.3 posts frontier-level coding scores, and the release notes flag emergent offensive-security capability the lab says it had to evaluate and gate before shipping.
Z.ai's GLM-5.3 posts frontier-level coding scores, and the release notes flag emergent offensive-security capability the lab says it had to evaluate and gate before shipping.
Bankers floating an Anthropic IPO are working from revenue growth steep enough to make a $2T listing plausible, which would be the largest debut ever.
Geoffrey Litt argues that once a model writes code faster than you can read it, the limiting factor stops being production and becomes comprehension — and that most AI tooling optimizes the wrong side of that.
Fernández-Villaverde and Norrick note that humanity is likely below replacement fertility as of 2026 — a first — and that the collapse is concentrated in low- and middle-income countries rather than rich ones.