Software Engineering & Architecture A new library of 177 production-ready Claude skills published this week highlights how AI coding workflows are maturing beyond individual tools: the same skill definitions run natively across Claude Code, OpenAI Codex, Gemini CLI, Cursor, Aider, Windsurf, and at least six other environments. The release — spanning nine business domains and including 17 specialized agents — reflects a broader push by practitioners to build portable abstractions rather than betting on a single AI assistant vendor. BrightCoding
Global Macro & Markets S&P Global's May 2026 economic outlook cuts its full-year global GDP growth forecast to 2.2% — down sharply from the 2.9% February estimate — with Brent crude expected to stay above $100/barrel through year-end as the Middle East conflict drags on. US equity markets have held near record highs on resilient corporate earnings, but the divergence between tech and the rest of the S&P 500 remains pronounced, and the ECB is now projected to add rate hikes that are expected to push several European economies into contraction. S&P Global
Enterprise AI & AI Strategy OpenAI published a detailed argument for what it calls the "next phase of enterprise AI," claiming the bottleneck has shifted from model access to deployment infrastructure and that most enterprises are still running disconnected point solutions rather than integrated agent workflows. The framing lands as Deloitte's 2026 AI report finds just 29% of organizations see meaningful ROI from generative AI and 95% of pilots still fail to reach production — underscoring that the gap between enterprise AI ambition and operational reality hasn't closed. OpenAI
AI Tools & Developer Utilities Anthropic updated Claude Managed Agents this month with three significant features: "Dreaming" (a scheduled process that reviews session history to improve agent memory over time), parallel multi-agent orchestration where a lead agent delegates to specialist sub-agents sharing a file system, and Outcomes (a rubric-graded loop that sends agents back to revise until a quality bar is met). Together these ship the scaffolding for long-running, self-correcting agent workflows without requiring developers to host their own orchestration infrastructure. 9to5Mac