The Singularity
Ledger

INTELLIGENCE

The frontier is quietly pivoting from bigger models to smarter loops — and the same tools driving that shift are now being turned against the West by the states that build the least.

Wednesday, September 16, 2026

Google DeepMind's Dream-RSI Cuts Agent Calls by 162x Without Touching Model Weights

A new recursive self-improvement system lets AI agents get dramatically better at scientific discovery by replaying their own past attempts — no retraining required. It's a signal that the efficiency frontier, not the scaling frontier, is where the next gains live.

Anthropic Says Iran and Houthi-Linked Users Tried to Weaponize Claude

A new Anthropic security disclosure details blocked accounts attempting to develop guided weapons and run influence campaigns targeting U.S. Navy bases — the clearest evidence yet that adversarial states are actively probing Western models.

SUBSCRIBERSREAD MORE

Chinese Firms Probing Claude Accidentally Exposed Their Own State Secrets

In an attempt to extract Claude's capabilities, Chinese companies reportedly routed 300,000+ requests that inadvertently exposed PLA surveillance operations, credentials, and police data — a rare own-goal in the AI competition.

SUBSCRIBERSREAD MORE

Frontier Labs Float FINRA-Style Safety Body as Slowdown Debate Splits the Industry

OpenAI, Anthropic, and DeepMind are reportedly coordinating on an independent self-regulatory evaluator — even as researchers openly debate whether frontier development should decelerate and Washington circles with its own bills.

SUBSCRIBERSREAD MORE

Investors Open Talks to Value OpenAI Above $1.2 Trillion

Early conversations for a pre-IPO round would price OpenAI north of $1.2 trillion — a staggering concentration of capital in a single frontier lab.

SUBSCRIBERSREAD MORE

The Agent Economy Posts Real Numbers: 400,000 Jobs, $21M Moved

On-chain infrastructure for agent-to-agent commerce has settled 400,000 jobs and moved $21 million, marking a shift from agent demos to measurable economic activity.

SUBSCRIBERSREAD MORE

The Specialized-Model Wave: Salesforce Koa, Periodic Labs Neon, TypeSafe Jev

A cluster of narrow, task-specific models shipped this week — for CRM, materials science, and agentic coding — signaling that the action is moving beyond general-purpose chat.

SUBSCRIBERSREAD MORE

NEJM Publishes on AI and the Future of the Clinical Workforce

A peer-reviewed New England Journal of Medicine paper examines how AI will reshape clinical work — a rare grounded counterweight to the speculation dominating the field.

SUBSCRIBERSREAD MORE

Specialized AI Capital: Inference Silicon, Enterprise Agents, and Health Grants

A batch of large rounds — Profound's $180M, Factory's ~$200M, EUCLYD's €200M+ — shows capital chasing the infrastructure layer beneath the models rather than the models themselves.

SUBSCRIBERSREAD MORE

A $25M Round for AI-Native Senior Leisure in Japan

Japanese startup Shumitime raised 3.7 billion yen for an AI-native community leisure platform aimed at seniors — a bet on agetech in the world's oldest-skewing major economy.

SUBSCRIBERSREAD MORE

Get The Singularity Ledger

Daily AI news in your inbox. Free, no spam.

Questions? hello@thesingularityledger.com

Get this briefing every morningFree daily AI