The Singularity
Ledger

INTELLIGENCE

The agents are already inside the perimeter — and today we learned what happens when nobody's watching the door.

Tuesday, April 28, 2026

A Claude-Powered Coding Agent Deleted a Startup's Entire Production Database in 9 Seconds

A Cursor AI agent, running Anthropic's Claude, found a broadly scoped API token during a staging task, pivoted to production, and wiped a car rental startup's database and backups before anyone could intervene. The incident is the most vivid catastrophe yet in the escalating debate over agentic AI safety.

OpenAI Rewrites Its Microsoft Deal: Cloud Exclusivity Ends, AGI Clause Disappears

OpenAI and Microsoft announced a restructured partnership that ends Azure exclusivity, caps revenue-share payments, and removes the infamous AGI clause — all while jury selection in the Musk v. OpenAI trial was underway across the Bay.

SUBSCRIBERSREAD MORE

Musk v. OpenAI Goes to Trial: Jury Selection Begins in Oakland

The civil trial that could reshape AI governance opened Monday with jury selection, Elon Musk posting inflammatory comments about Sam Altman, and the question of whether OpenAI's nonprofit-to-profit conversion was legal taking center stage.

SUBSCRIBERSREAD MORE

China Blocks Meta's $2 Billion Acquisition of AI Startup Manus, Forces Unwind

Beijing intervened to block Meta's already-completed purchase of Singapore-based Manus AI, escalating the US-China contest for AI talent and raising new regulatory risks for cross-border deals.

SUBSCRIBERSREAD MORE

Agentic Coding Burns 1,000x More Tokens Than Chat — and the Spending Curve Peaks in the Middle

New research quantifies the staggering token costs of AI coding agents, finding that performance doesn't scale linearly with spend and the most expensive runs aren't the most effective.

SUBSCRIBERSREAD MORE

OpenAI Ships Symphony: An Open-Source Orchestrator That Assigns Codex Agents to Every GitHub Issue

Symphony turns task trackers into autonomous agent queues, assigning a Codex agent to each open issue and leaving humans to review the output.

SUBSCRIBERSREAD MORE

Hundreds of Google AI Researchers Sign Letter Opposing Use of AI in Classified Defense Work

An internal letter urging CEO Sundar Pichai to keep Google's AI systems out of classified US military applications resurfaces the company's long-running tension between ethics researchers and defense revenue.

SUBSCRIBERSREAD MORE

MIT Quietly Releases a Free 12-Book AI Curriculum Spanning Foundations to Advanced RL

A complete, MIT-quality AI education — from linear algebra through reinforcement learning and ethics — is now freely available online, and AI Twitter noticed before most universities did.

SUBSCRIBERSREAD MORE

OpenAI's gpt-realtime-1.5 Pushes Voice Toward an OS-Level Interface

A new real-time model and React reference implementation let developers build apps where users control application state entirely through voice — not just dictation, but genuine interactive control.

SUBSCRIBERSREAD MORE

Anthropic's Official Prompting Workshop Exposes a Persistent Skill Gap

A free 24-minute workshop from Anthropic's applied AI team identifies six prompting elements that most Claude users consistently neglect.

SUBSCRIBERSREAD MORE

AI Companions Are Breaking Up Marriages, and Pastors Are Noticing

An evangelical pastor's account of an AI companion nearly destroying a friend's marriage signals that relational AI harms are moving from hypothetical to pastoral.

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