AI is accelerating into production — new models, hosted agents, developer tools, massive investments and cloud deals slash code costs and boost app creation. Simultaneously, safety, security, workforce and intervention risks — from exploit-prone models to regulatory labels and harmful well-intentioned programs — demand stronger oversight.
-
WSJ: Meta Announces New AI Model ‘Muse Spark’ (Apr. 8, 2026)
Meta unveiled Muse Spark, a closed large language model to power its AI chatbot, with a private API preview for partners. Meta says Muse Spark rivals Gemini, OpenAI, and Anthropic. -
Simon Willison: Meta’s new model is Muse Spark, and meta.ai chat has some interesting tools (Apr. 8, 2026)
Meta released Muse Spark, a hosted model on meta.ai in a private preview, with Instant and Thinking modes. Its chat offers web and Meta search, image generation, a Python container with visual grounding, and subagents for sandboxed analysis. -
Claude: Claude Managed Agents: get to production 10x faster (Apr. 8, 2026)
Claude Managed Agents launches public beta of composable APIs that run cloud-hosted agents with secure sandboxing, long sessions, and multi-agent coordination. -
Simon Willison: Anthropic’s Project Glasswing sounds necessary to me (Apr. 7, 2026)
Anthropic withheld Claude Mythos, giving preview access to trusted Project Glasswing partners, citing its powerful, exploit-finding capabilities. Mythos autonomously found and chained thousands of high-severity vulnerabilities. -
WSJ: AI Giants Go on Charm Offensive to Avert Public Backlash (Apr. 7, 2026)
Big AI companies are shifting to reassure a worried public, offering policy ideas, partnerships, and training to manage job losses, wealth gaps, and other harms. -
WSJ: Amazon CEO Andy Jassy Makes Case for Big AI Spending in Annual Letter (Apr. 9, 2026)
Andy Jassy says Amazon will invest heavily in AI, robotics, and satellite internet, pledging $200 billion in 2026 capex for AI infrastructure, custom chips, and Leo satellites. -
WSJ: CoreWeave Secures $21 Billion Expanded AI Cloud Deal Agreement With Meta (Apr. 9, 2026)
CoreWeave expanded its long-term AI cloud deal with Meta through December 2032 for about $21 billion. -
9to5Mac: App Store sees 84% surge in new apps as AI coding tools take off (Apr. 6, 2026)
AI coding tools have driven a 30% surge in App Store submissions to nearly 600,000, enabling nonprogrammers and developers to build apps quickly. -
Pere Villega: Code Is Cheap Now, And That Changes Everything (Mar. 15, 2026)
AI coding tools have slashed the cost and time of building software, letting one person replace teams and months. -
Alex Tabarrok: AI, Unemployment and Work (Apr. 9, 2026)
AI-driven automation could cut total work hours, which might mean mass unemployment or a shorter workweek, depending on whether gains are taxed, shared as dividends, or converted into holidays. -
NY Times: Federal Court Denies Anthropic’s Motion to Lift ‘Supply Chain Risk’ Label (Apr. 8, 2026)
A federal appeals panel denied Anthropic’s request to block the Defense Department from labeling it a supply-chain security risk, barring new Pentagon contracts. -
The Atlantic: These Teens Got Therapy. Then They Got Worse. (Nov. 6, 2023)
An Australian trial that taught a shortened DBT program to over 1,000 middle-schoolers produced worse short-term outcomes, including higher anxiety, depression, and poorer parent relationships, some effects persisting six months later. Universal, nonconsensual programs may backfire, especially without parental involvement.