AI news6
Ollama adds macOS speech and four sizes of decision model
Ollama’s stable 0.40.11 release, published 24 October, adds streamed speech generation on macOS. It also introduces d3 decision models in 2B, 4B, 9B and 27B sizes for classification and scoring.
github.com ↗Why it matters
This new local release extends deployment options beyond the earlier vLLM inference and Baseten decision-model stories.
Axon announces Body 5 with 25 times the AI processing power
Axon’s 24 October announcement claims 25 times the AI processing power and 20% more battery life than earlier cameras. Its Guardian conflict alerts require human review; general availability dates remain unannounced.
axon.com ↗Why it matters
Unlike the earlier Google and Microsoft Guardian products, this is AI moving into police field equipment, where oversight has immediate consequences.
Virginia orders a statewide plan for AI-related cyberattacks
Virginia’s 23 October Executive Directive 6 orders a cyberattack prevention and response plan and a Cyber Consortium. It also calls for regular briefings with AI companies, researchers and regulators.
governor.virginia.gov ↗Why it matters
This new state directive differs from earlier Virginia grants and tutoring pilots by assigning a coordinated public-sector response to agent risks.
Meta acquires AI analytics startup Sundial
Sundial co-founder Chandra Narayanan announced its acquisition by Meta on 23 October. He and Julie Zhuo are returning with team members to build AI analytics tools and agents.
linkedin.com ↗Why it matters
This acquisition adds analytics expertise, a separate development from Meta’s earlier model releases and TPU work.
SAP CTO reports more than 60,000 internal Joule Work users
In a sponsored interview published 25 October, SAP CTO Philipp Herzig says more than 60,000 colleagues use Joule Work internally. He identifies adoption as harder than the technology and recommends keeping models replaceable.
therundown.ai ↗Why it matters
Unlike Friday’s SAP reranker benchmark, this new adoption figure highlights the organisational work involved in scaling agents.
NIH adopts “Super Intelligence” terminology in official communications
An NIH notice dated 23 October says federal agencies are transitioning from “AI” to “Super Intelligence” in official documents and communications. It retains a definition covering machine predictions, recommendations and decisions.
grants.nih.gov ↗Why it matters
This agency implementation notice follows the previously examined federal naming order; terminology alone establishes no new model capability.