Software Engineering & Architecture
Microsoft's Work IQ APIs reached general availability today, giving agent builders a unified intelligence layer over Microsoft 365 data — email, calendar, meetings, files, people, and org-graph relationships — exposed through A2A, MCP, and REST protocols in a single consumption-billed surface. The design treats organizational context as infrastructure: agents connect to Work IQ rather than scraping individual M365 endpoints, and digital workspaces persist intermediate outputs and memory within the tenant boundary for multi-step tasks. For teams building on GitHub Copilot, Foundry, or Copilot Studio, the shared context layer eliminates redundant graph queries and enables agents to reason about who works with whom without bespoke integrations. Microsoft 365 Dev Blog
Global Macro & Markets
The Bank of Japan raised its benchmark rate by 25bp to 1% today — its highest since 1995 — even as Governor Ueda participated only in writing due to hospitalization for a hepatic cyst infection. The move, anticipated by 94% of economists surveyed by Reuters, was framed around persistent inflation upside from the Iran war-driven energy shock, and officials signaled scope for further hikes given still-negative real interest rates. The decision is the clearest signal yet that Japan's post-deflation normalization has institutional momentum independent of any single policymaker. Reuters
Enterprise AI & AI Strategy
A new Writer survey of over 1,600 enterprise AI decision-makers finds that 79% report significant challenges scaling AI despite high investment — and only 29% see meaningful ROI from generative AI, with agent deployments faring even worse at 23%. The report identifies the gap between pilot success and production reliability as the central blocker: teams can demo agents that work but struggle to make them work consistently at scale across real enterprise data and workflows. The finding lands as vendors like NTT DATA and Google are betting large on closing exactly this gap through structured deployment programs. Writer
AI Tools & Developer Utilities
MiniMax released M3 on June 1, but the open weights and technical report landed this week as promised — making it the first open-weight model with frontier-tier software engineering scores (59% on SWE-Bench Pro, ahead of GPT-5.5), a 1-million-token context window, and native multimodal computer use in a single package. The new MiniMax Sparse Attention architecture delivers roughly 15x faster decoding at 1M context versus its predecessor, and API pricing at $0.30/$1.20 per million tokens makes it the most cost-competitive frontier coding model available. For teams that need long-context agentic coding without sending data to closed APIs, M3 changes the calculus. DataNorth