Home

Recent college graduates have faced a tougher job market in recent years, with unemployment rising relative to the broader workforce. In summer 2026, however, there was no major increase in unemployment among 22- to 25-year-olds with bachelor’s degrees who had left school. Their unemployment rate averaged 7.3%, which was higher than in 2022, 2023, and 2025 but lower than the 7.8% recorded in 2024. Unemployment was also only slightly higher in occupations where artificial intelligence can be used extensively, and the difference was not statistically significant. Recent graduates did not lose significant ground compared with older college graduates or similarly aged workers without college degrees. A broader measure that included graduates who wanted jobs but had stopped actively searching reached 10.4%, the highest level in the period studied, although this increase was also not statistically significant. At the same time, other research suggests that graduates in majors highly exposed to AI, especially computer science, have been slower to find their first jobs and may be entering lower-paying industries such as retail and restaurants. Remote work may also be contributing to the difficulty because younger workers can benefit from in-person mentoring, feedback, and informal learning, although remote-friendly jobs are often also more exposed to AI. Overall, the summer of 2026 provides little evidence of a widespread AI-driven unemployment crisis for new college graduates, but the effects may become clearer as AI becomes more deeply integrated into workplaces.
https://anderson-review.ucla.edu/for-new-college-grads-the-ai-jobpocalypse-didnt-arrive

Microsoft Fabric Graph now supports unbounded variable-length paths, allowing queries to follow relationships as deeply as needed instead of being limited to eight hops. This makes it easier to analyze complex chains such as conversations, data lineage, dependencies, organizational hierarchies, and supply chains, while bounded paths remain preferable when depth, performance, or cost needs to be controlled.

https://www.linkedin.com/pulse/explore-deeper-connections-unbounded-variable-length-paths-fine-9oohc/

Imperial College London recently convened researchers and industry experts to explore how nature-inspired computing could help meet the surging demand for AI while addressing the unsustainable energy and infrastructure costs of current systems. Drawing on biological models like the human brain and plant networks, new hardware paradigms such as neuromorphic chips aim to deliver major gains in efficiency and performance, but real-world adoption hinges on overcoming barriers in skills, manufacturing scale-up, investment, and regulation. Success will depend on converging disciplines and sectors—universities, startups, and large tech firms working together—to co-design, demonstrate, and deploy these next-generation technologies.
https://www.imperial.ac.uk/news/articles/convergence-science/2026/nature-inspired-computation-real-world-challenges-and-opportunities-/

Two new studies highlight significant gaps between the promises of technology and the real-world experiences of users, particularly regarding social media safety communications and AI-driven health advice. Research on major platforms reveals that safety features are often described vaguely or selectively, shifting responsibility onto families rather than addressing underlying design and business practices. Meanwhile, an analysis of Alzheimer’s caregiving support shows that while AI chatbots sound polite and informative, they lack the contextual depth, emotional nuance, and lived experience that human caregivers and clinicians provide.

https://icsi.berkeley.edu/gaps-in-social-media-risk-communication-and-ai-health-advice/

Microsoft Fabric Graph has added support for additional #GQL features with richer path semantics, traversal controls, and composable query constructs for analyzing complex graph structures. Variable-length and unbounded paths can be combined with modes such as WALK, TRAIL, SIMPLE, ACYCLIC, and DIFFERENT EDGES to control whether vertices or edges may be revisited during traversal. ANY SHORTEST provides shortest-path matching, while functions such as NODES, EDGES, INNER_NODES, SOURCE, DESTINATION, and ELEMENT_ID expose individual components and identifiers of matched paths. Query composition is enhanced through NEXT, EXISTS, CALL, OPTIONAL CALL, and UNION DISTINCT, enabling multi-stage pattern matching, conditional execution, subqueries, and deduplicated result sets. Aggregation is more expressive through filtered and nested aggregate expressions, alongside COLLECT_ONE, COLLECT_LIST, and COLLECT_ELEMENTS for constructing scalar and collection-based results from graph matches. The language also supports dynamic nested values and explicit type conversion with CAST, together with functions such as RANGE and POWER for generating sequences and performing numerical operations within queries.

https://learn.microsoft.com/en-us/fabric/graph/gql-language-guide