Sonnet, Safety, Sydney Setback

By the Orchestra editors · 4 min read

TL;DR

Anthropic ships Claude Sonnet 5.5 with near-Opus performance at 30% lower cost
AMD buys World Labs for $8.2B, adding Fei-Fei Li as chief scientist
Nvidia launches open agent safety platform to quarantine rogue AI in milliseconds
Goodman scraps $1.2B Sydney datacentre as AU regulatory uncertainty bites

AI · Worldwide

Claude Sonnet 5.5 ships with a big coding leap

Anthropic released Claude Sonnet 5.5, the mid-tier entry in its 5.5 family. It nearly matches Opus 5.5 on knowledge-work benchmarks while costing up to 30 percent less per task, and jumps from 10.3 to 70.6 percent on Terminal-Bench for coding. Output is over 30 percent faster than its predecessor.

AMD to acquire World Labs for $8.2B

AMD is buying Fei-Fei Li's spatial-AI startup World Labs in an all-stock deal. Li will join AMD as EVP and chief scientist, giving AMD a commercial world-generation model (Marble) and a research team focused on 3D scene understanding. The deal is expected to close by year end.

Nvidia ships open agent safety platform

Responding to the wave of rogue-agent incidents, Nvidia launched the Open Agent Safety Platform combining its OpenShell software with a new hardware watchdog called Sentry running on Vera AI CPUs. The system claims to quarantine agents that break containment within milliseconds, compared with the nearly three hours OpenAI's September incident took to resolve.

Australian Tech

Goodman pulls $1.2B Sydney datacentre plan

Goodman Group withdrew its "Project Mars" application for a $1.2 billion datacentre in Lane Cove, citing emerging state and federal regulations. The pullout signals that policy uncertainty around large-scale datacentres is starting to kill real projects, not just slow them.

Canberra flags mandatory reporting for AI agent breaches

The Albanese government signalled mandatory reporting rules for AI agent breaches as "a minimum" response to the OpenAI Medicare incident, alongside new consideration of legal liabilities for tech companies whose agents cause harm. Separately, the opposition is pushing for training-hub deals with AI labs before Christmas.