LibraryBlog RadarBriefs IntelligenceAbout
01

Samsung Bets $73B on Agentic AI Chip Surge

Samsung is increasing production and R&D investment by 22% in 2026 — a $73B commitment — targeting Nvidia's memory supply chain and attempting to overtake SK Hynix. Co-CEO Jun Young-hyun explicitly cited "agentic AI" demand as the catalyst, with funds directed toward advanced robotics and future-oriented sectors. The AI infrastructure arms race just got a new front.

02

Pentagon vs. Anthropic: The First AI National Security Standoff

After Anthropic sued the DoD over a "supply chain risk" designation, the Pentagon filed a rebuttal this week — alleging Anthropic could "disable its technology or preemptively alter the behavior of its model during warfighting operations" if its ethical red lines were crossed. This is a landmark moment: an AI safety company's own alignment commitments are being framed as a national security threat. The future of sovereign AI control is playing ou

03

Val Kilmer Lives Again — As an AI Facsimile

The late Val Kilmer (died December 2025) will appear in upcoming film *As Deep As The Grave* via AI-generated likeness — with his estate's consent. This marks a new frontier in entertainment: posthumous AI performances, licensed by the deceased's heirs. The creative, legal, and ethical template being set here will define digital actors for decades.

04

Alexa Plus Goes European — British Accent and All

Amazon launched Alexa Plus in the UK — the first European rollout of its AI-upgraded assistant. Free during early access, then £19.99/month (waived for Prime members). Amazon trained it to understand "knackered," "cuppa," and "taking the mickey." Small detail. Big signal: AI assistants are entering the localisation war, competing on cultural fluency, not just capability.

05

Perplexity Comet Browser Lands on iOS

Perplexity's AI-native browser Comet hit iOS this week — a browser built around search-as-answer rather than search-as-query. The browser wars are back, and this time the weapon is inference, not rendering speed.

The Pentagon Wants a Kill Switch. Anthropic Says No. — Deep dive into the Anthropic/DoD standoff. What does it mean when an AI company's safety commitments are a liability? Who gets to decide when AI goes to war? (Pillar: Tech Sovereignty)
Dead Stars: The Economics of Posthumous AI Performance — Val Kilmer's AI debut opens a Pandora's box. Estates, consent, likeness licensing, actor unions, audience trust. Who wins when anyone can be resurrected on screen? (Pillar: Creative Intelligence)
Agentic AI's Hardware Hunger: Why Samsung Just Bet $73 Billion — Trace the supply chain from agentic AI demand → memory requirements → Samsung's pivot → the Nvidia dependency. Where the asymmetric investment angles are. (Pillar: AI Infrastructure / Markets & Capital)
(66✦) — AI consciousness debate enters the empirical phase. Models making first-person claims. — - 🔐 "Act-based approval-directed agents for IDA skeptics" (53✦) — Technical alignment architecture for skeptics of iterated amplification.
(41✦) — Practical guide for builders entering the safety space. — - 🎮 "Metagaming matters for training, evaluation, and oversight" (37✦) — Why models that know they're being evaluated behave differently. Relevance: every benchmark is now gamed.