The Handover
Wednesday 7 October 2026
The day in 7 stories, as Asia signs off and New York wakes up.
Australia
Landmark High Court ruling requires climate consideration for fossil fuel approvals
Australia's High Court ruled 3-2 that fossil fuel approvals must consider local environmental effects from emissions, including coal burned overseas. The decision sets a binding precedent for resource approvals nationwide.
AI
Mistral releases 'Le Chonk,' its largest open-weights AI model yet, with a focus on cybersecurity and niche domains
Mistral unveiled Mistral Large 4, nicknamed Le Chonk, a trillion-parameter open weights model trained entirely in European data centres. It is aimed at sovereign AI and cybersecurity as a rival to US and Chinese systems.
Australia
Anthropic given draft data centre rules before publication, FOI documents show
The Australian government gave Anthropic draft data centre guidelines five days before publication while negotiating its agreement to follow them. Briefing notes say it was consulted on the rules; Anthropic denies influencing the final text.
AI
OpenAI releases solutions to hundreds of open mathematics problems
OpenAI released 722 manuscripts from an unreleased frontier model, detailing solutions to long-standing mathematics problems in 372 result families. Independent mathematicians' group AGMAI guided the release.
Tech
Google's Spanner Omni Reaches General Availability, Replacing Proprietary Hardware with Software
Google made Spanner Omni generally available, letting its distributed SQL database run in customer data centres, other clouds or on a laptop. It replaces Colossus and TrueTime, but comes without an availability SLA.
AI
Meta, Walmart and Stripe launch protocol to standardise AI agent interactions
Meta, Walmart, Stripe and other companies published the open Personal Agent Protocol to govern how AI agents interact with businesses. It seeks to standardise access as retailers resist unauthorised scraping by Meta's Muse.
AI research
MEND selectively moves flow-model samples when reward gains justify distance
University of Texas at Austin and Google researchers introduced MEND, which updates flow-model samples only when reward gains exceed a quadratic movement cost. It improved several benchmarks faster than Flow-GRPO, ReFL and DiffusionNFT.
The Handover is the day in a handful of stories, weekday evenings at 8pm your time, and a thread on X at 7am in New York and in London. Get it by email or push from your account, or subscribe by email.
Also posted as a thread on X: @zotpaper.