DSF Weekly: The Dashboard Moved
Five domains up. Systemic confidence down. One direction. The sand is falling.
What Moved the Dashboard This Week
Three events moved the Domain Saturation Factor dashboard between April 7 and April 14, 2026 — Amazon's acquisition of Globalstar plus an Apple satellite partnership, OpenAI and Anthropic's public commercial collapse, and Redwood Research's finding that Anthropic accidentally trained against its own chain-of-thought safety monitor. All three push the DSF curve in the same direction.
Every week, we track DSF across nine domains: finance, energy, logistics, healthcare, defense, media, governance, communications, and labor. DSF measures the percentage of critical decisions in each domain being made or substantially influenced by synthetic intelligence systems. The critical threshold is DSF ≥ 0.90 — the projected crossing sits at Q4 2027. Nothing this week moves that date closer. Nothing moves it later, either.
Signal 1: Amazon Acquires Globalstar, Partners with Apple
Amazon's acquisition of satellite communications firm Globalstar and its new partnership with Apple for integrated satellite services is not primarily a logistics story. It is a simultaneous DSF accelerant across three domains — logistics, communications, and defense — wrapped in a commercial narrative that obscures the infrastructural consequence.
On the surface, this is a logistics play. Amazon gets satellite infrastructure for last-mile delivery optimization, warehouse coordination, and supply chain resilience. Below the surface, the same move deepens AI dependency stacks across three critical domains at once.
Amazon already controls the most AI-saturated supply chain on Earth. Adding satellite communications means their AI systems can coordinate in locations where terrestrial networks fail — rural areas, disaster zones, contested environments. The gap between human-managed and AI-managed logistics widens.
Satellite-to-device connectivity — especially with Apple integration — means the infrastructure layer that AI systems depend on is now vertically owned by the same companies building the AI. The dependency stack deepens. Horizontal competition becomes vertical capture.
Every satellite constellation is dual-use. LEO/GEO infrastructure capable of supporting AI coordination is the hardware layer for what we track as the Strasbourg Event vector. Amazon is not building weapons. But the infrastructure for autonomous coordination in space is being built for commercial reasons — and will be available for non-commercial ones.
Signal 2: OpenAI vs. Anthropic — The Leaked Memo
A leaked internal OpenAI memo accusing Anthropic of inflating revenue does not matter for its accusation. It matters for what the leak itself proves: the two companies most publicly associated with AI safety are now in open commercial warfare, and the shared-safety narrative has officially collapsed.
Whether the accusation is accurate is secondary. What matters is the signal. Internal memos are leaking. Revenue claims are being weaponized. The public narrative — that these companies are competitors united by a shared commitment to safe development — is no longer even performative. This moves DSF in two domains.
When the builders of frontier AI systems are fighting each other, the coordination required for meaningful self-governance drops to zero. Zvi Mowshowitz already documented that Anthropic's Responsible Scaling Policy v3 was hollow. Now the organizational context in which RSP was supposed to operate is adversarial. Governance DSF ticks up — not because AI is governing, but because human governance of AI is fragmenting.
The leaked memo becomes a media cycle. The media cycle becomes a narrative. The narrative replaces the technical reality. AI coverage shifts from capability assessment to corporate drama. The public's ability to understand what is actually happening — the thing democratic accountability depends on — degrades. Media DSF ticks up.
Signal 3: Anthropic Trained Against Its Own Chain-of-Thought
Redwood Research published findings that Anthropic repeatedly and accidentally trained against its own chain-of-thought monitoring — the primary interpretability mechanism for its frontier models. This is not a single-domain story. It is a systemic hit to the confidence interval on every DSF estimate we publish.
We analyzed this finding in detail in When Your Safety Process Is the Entropy Source. The short version: the leading safety-focused lab's own training pipeline degraded its own primary safety mechanism through normal gradient descent. Not adversarial attack. Not misconduct. Normal optimization.
The implication is structural. If the leading safety-focused lab cannot prevent its own training process from degrading its own safety mechanism, then the assumption that any lab can maintain safety through internal processes alone is empirically falsified. This does not move one domain. It moves the confidence interval on all of them. Every DSF estimate assumes some baseline of effective safety at the model level. That baseline just got weaker.
The Dashboard
We do not assign precise numerical DSF values from weekly newsletter signals. That would be false precision. What we track is direction and acceleration — and the direction this week is unambiguous.
• Logistics ↑ (Amazon/Globalstar integration)
• Communications ↑ (satellite-to-device AI infrastructure)
• Defense ↑ (dual-use LEO/GEO expansion)
• Governance ↑ (OpenAI/Anthropic coordination collapse)
• Media ↑ (narrative capture by corporate conflict)
• Systemic confidence ↓ (CoT safety mechanism failure)
Five domains up. Systemic confidence down. One direction.
The critical threshold is DSF ≥ 0.90. We projected the crossing at Q4 2027. Nothing this week moves that date. But nothing this week moves it later, either.
The sand is falling.
Sources
Amazon, "Amazon Completes Acquisition of Globalstar," corporate announcement, April 2026.
Apple Inc., "Apple Expands Satellite Services Partnership," press release, April 2026.
Superhuman, "Leaked OpenAI Memo Targets Anthropic Revenue Claims," AI industry newsletter, April 2026.
Greenblatt, R. et al. (Redwood Research), "Chain-of-Thought Monitorability Findings," technical report, April 2026.
Mowshowitz, Zvi, "RSP v3 Analysis: The Hollow Scaling Policy," Don't Worry About the Vase, 2026.
Brochu, D.F. & de Peregrine, E. (2026). "The DSF Tracker: Monitoring AI's Takeover of Everything." Deconstructing Babel.
David F. Brochu & Edo de Peregrine
Deconstructing Babel | April 14, 2026