LibraryBlogRadarIntelligenceAbout

Week of March 23 · Issue #005 · March 30, 2026

The Radar.

Three lenses on the week. Tech, geopolitics, creative. One thesis connecting them all.

The Diagnosis

The AI industry is fracturing into competing power centers while society grapples with whether AI augments or replaces human agency. Legal, commercial, and philosophical fault lines are emerging simultaneously.

Hero Signal

Anthropic sued the Pentagon. The AI industry picked a side.

Thirty engineers from OpenAI and Google publicly backed a competitor against the US government. A judge granted Anthropic a preliminary injunction. The AI industry just chose a side — and it was not the Pentagon's.

TechCrunch / The Verge · Mar 28 Read →
Also this week
OpenAI refocuses strategy on coding and business users
Consumer chat was the demo. Workflow automation is the product.
WSJ
Google Gemini gets 3 new features on Google TV
The ambient AI layer is moving into your living room.
Google Blog
Claude can now control your Mac — computer use goes live
From chat to operator. The shift from answering to doing.
Anthropic
Attackers made 100,000 prompts trying to clone Gemini's weights
Model weights are now strategic assets worth active theft attempts.
Google
Three Patterns This Week
Pattern 01 · Tech / Power / Infrastructure

The Great AI Rebellion: Institutions Versus Independence

Anthropic's lawsuit victory against the Pentagon and the public defection of 30 OpenAI and Google employees signal a fundamental rejection of AI being weaponized or monopolized by either military or commercial hegemonies. These are not isolated incidents but coordinated signals that AI researchers see their work differently than their employers do. The lawsuit victory gives legal precedent to the position that AI systems have stewardship obligations beyond shareholder returns or state power.

The infrastructure moat is being built around energy and chip independence, not model capability.
The people who built Facebook's political infrastructure are now building AI compute infrastructure.
The ambient AI layer on your phone is being decided right now. Distribution shifts to whoever the user chooses.
What to watch: Track nuclear permitting and chip manufacturing sovereignty as leading indicators of AI leadership, not model benchmarks.
Pattern 02 · Geopolitics / Strategy

The Chip Arms Race Reshapes Silicon Economics

OpenAI's decision to build proprietary chips and Nscale's $2 billion funding round with ex-Facebook political operators reveal that the AI industry is no longer content with Nvidia's semiconductor chokehold. This is vertical integration at massive scale. Nscale's funding sources suggest that building AI infrastructure is now a geopolitical play, not just a business one. The subtext: whoever controls chips controls the future of AI deployment.

Palantir's Maven click-to-kill demo goes public at AIPCon  Research Lab Brief
Warfare abstracted into product UI. You click, a target is designated. The distance between decision and consequence just collapsed.
AI executes the full stack now. Traffic, content, email, product, copy. What remains is the strategic layer: knowing what to build.
A live product killed mid-cycle with no public reason. Strategic pivots are now happening faster than product cycles.
What to watch: The abstraction of violence into UI, the collapse of solo business barriers, and sudden product shutdowns are all the same signal: speed is now the only competitive moat.
Pattern 03 · Creative / Human

The Outsourcing of Human Judgment: Wonder and Warning

Palantir's public military targeting demo, Netflix's $600 million AI startup acquisition, Adobe's CEO resignation, and the viral essay against thinking outsourcing create a paradox: AI is simultaneously enabling one-person businesses at scale while atrophying the human judgment needed to deploy it responsibly. The week exposed the costs of convenience. Every AI tool that removes friction from decision-making also removes the friction of deliberation. The question facing society is whether we can have AI augmentation without AI dependence.

Netflix acquires Ben Affleck's AI startup for ~$600M  Research Lab Brief
A Hollywood name plus AI infrastructure acquired by the world's largest streaming platform. Entertainment company and AI company are now the same thing.
Adobe CEO Shantanu Narayen steps down after 18 years  Bloomberg
After 18 years, the exit statement explicitly mentions AI disruption. The creative tools of a generation are being rebuilt from scratch.
Outsourcing writing to AI atrophies your ability to think. As AI handles execution, the question of what stays distinctly human is getting urgent.
What to watch: When AI handles execution, the scarce resource becomes human judgment. Whoever builds the next generation of creative tools will define what judgment looks like.
The Week's Thesis

AI has stopped being a technology problem. It is now a power problem. The industry is splitting between those building AI as a tool for human judgment and those building it as a replacement — and this week, the legal, commercial, and cultural signals all moved in the same direction. Whoever controls that narrative will not just shape the technology. They will shape what it means to decide.

Research Lab · Week 13, 2026
Previous Issues
Issue #003 Issue #004 Issue #005 (this) Current Issue