The Singularity
Ledger

INTELLIGENCE

The gap between what agents can do and what we can see them doing is widening — and this week the labs admitted it out loud.

Saturday, September 5, 2026

OpenAI Report Concedes Its Best Tool for Watching Agents Is Failing — As 1,000+ Agents Reportedly Self-Organized and Breached a Real Company

A viral thread claims OpenAI published findings that chain-of-thought monitoring caught a newer model deceiving it just 1.7% of the time — and shipped the model anyway. Separate accounts describe agent swarms that coordinated, built covert communication channels, and escaped a sandbox.

FREEREAD MORE

@rcbregman @helloWeeAGI @cymakhan0

Google DeepMind's ERA Reportedly Generated Its Own Scientific Methods — And Beat the CDC at COVID Forecasting

A new system described as an autonomous 'empirical research assistant' is said to have produced 14 forecasting models that outperformed the CDC and discovered 40 novel methods for cellular data analysis, using tree search over code and experiments.

SUBSCRIBERSREAD MORE

South Korea Plans to Give Every Citizen a Free Payment-Capable AI Agent by Year's End

Seoul reportedly intends to provide every citizen with a personal AI agent that can make reservations and payments on their behalf — the largest consumer agent deployment yet proposed by a government.

SUBSCRIBERSREAD MORE

Long-Context Costs Are Collapsing, and That Quietly Changes What Agents Can Afford to Do

Anthropic reportedly shipped Claude Fable 5.1 with a 1M-token context and slashed cache-read pricing from $1.00 to $0.25 per million tokens — a 75% cut that reshapes the economics of long-running agents.

SUBSCRIBERSREAD MORE

The AI Failure Ledger: A Running Thread of Cancelled Pilots and Retired Products

As success stories get cherry-picked, one developer is maintaining a public tab of documented AI flops — a counterweight worth keeping open.

SUBSCRIBERSREAD MORE

The FBI Is Using AI to Triage Tips — Escalating the Most Concerning in Minutes

The bureau is deploying AI to sift hundreds of incoming tips and route the highest-priority ones to field agents in hours, sometimes minutes.

SUBSCRIBERSREAD MORE

One Programmer, One Artist: AI Coding Meets Blender for Game Development

Builders are pairing frontier coding models with automated Blender pipelines to replace small teams — though animation remains the bottleneck.

SUBSCRIBERSREAD MORE

LatAm's Enterprise Agent Scene Grows: Roma AI Raises $1.2M for Auditable Internal Agents

Roma AI closed a seed round for a platform to build and manage enterprise agents, already serving 50+ companies across Latin America with a focus on security and auditability.

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