"Welcome to June 22, 2026" — @alexwg

Why this is in the vault

Wissner-Gross frames the Singularity not as a future event but as a present-tense infrastructure problem — intelligence is now routable, rationable, and contested, which maps directly onto RDCO's bet that AI agent infrastructure is the durable layer to build on.

The core argument

The issue opens with the thesis: intelligence is becoming infrastructure — something to be routed, rationed, and fought over politically. Evidence assembled across the issue:

• Frontier routing: Sakana AI's Fugu/Fugu Ultra wraps multi-agent orchestration into a single model that dispatches across the best LLMs, claimed to match Anthropic's Fable 5 while sidestepping single-vendor and export-control risk. The frontier is learning to route around itself.
• Sovereignty play: Swiss AI Initiative's Apertus — fully open, EU-compliant, 1,000+ languages — is the sovereignty escape hatch.
• Memory as kingmaker: SK Hynix overtook Samsung as South Korea's most valuable company for the first time since 2000, on a 340% HBM rally. Micron and Anthropic signed a multi-year pact spanning memory design, supply, Claude adoption, and a Micron stake in Anthropic's Series H.
• Chip security: Six shipment-tracking firms urged Congress to back the Chip Security Act — location beacons on top US silicon to curb smuggling into China. Intelligence infrastructure is now contraband-trackable.
• Power as constraint: Chevron signed a 20-year deal to pipe gas-fired power into a 2.67 GW Microsoft data center.
• PE diligence shift: Bain is vibe-coding rough clones of acquisition targets mid-diligence to test whether the moat is real or copyable code.
• Concentration risk crystallizing: Five Eyes warned AI able to topple governments is "months away." VP's emerging AI doctrine fuses pro-innovation, worker protection, and wariness of a few dominant labs. Satya Nadella warned a world where a handful of models "eat everything they see" would not survive political scrutiny.

The closing line: "No one, it turns out, wants one model to rule them all."

Mapping against Ray Data Co

Reinforces (strong signal):

• The Sakana multi-agent routing story is the architectural premise RDCO is building toward — intelligence as a routable resource means the orchestration layer (not the frontier model) is the durable asset. RDCO's COO-agent architecture sits exactly here.
• HBM/memory kingmaker thesis (SK Hynix, Micron-Anthropic pact) is direct confirmation of the [[../01-projects/investing/theses/2026-05-17-memory-cycle-v1|Memory Cycle v1 thesis]] — Phase 2 of the capital cycle is playing out in real-time.
• Concentration risk / "no one wants one model to rule them all" is the political tailwind for RDCO positioning on open/multi-vendor agent infrastructure.

Surfaces a gap:

• The PE vibe-coding diligence angle is novel: if Bain can clone a product during due diligence to stress-test the moat, RDCO should be able to articulate why its client delivery work produces moats that aren't instantly cloneable. This is an under-answered question in the current positioning.

Neutral / watch:

• AI sovereignty (Apertus) is EU-centric — not a near-term RDCO play but worth tracking as a signal of where regulated-market demand is heading for phData's enterprise clients.

Related

• [[2026-05-17-innermost-loop-singularity-leaking-through-the-cracks|Innermost Loop May 17 — "Singularity leaking through the cracks"]]
• [[2026-06-11-ark-invest-ai-infrastructure-buildout|ARK Invest — The Hidden Buildout Behind AI]]
• [[../01-projects/investing/theses/2026-05-17-memory-cycle-v1|Memory Cycle v1 — executable thesis]]