Semantic Kernel vs Gemini

Semantic Kernel's most recent Python release (1.42.0) added an explicit 'Microsoft Agent Framework successor callout' to the READMEs — Microsoft is publicly pointing users toward a different framework as the forward path. The rest of the recent cadence is consistent with a project in late-stage maintenance: security hardening (path validation in CloudDrivePlugin, gRPC plugin, OpenAPI plugin; SQL escaping in connectors; Snappier and Kiota vulnerability bumps), dependency bumps via dependabot, vector-store connector polish, and small prompt-template fixes. Feature additions are narrow — ImageContent in tool/function results, OpenAI text-to-image model support, prompt template serialization improvements.

The project is transitioning from active framework to maintained predecessor. Microsoft's agent stack is consolidating under the new Microsoft Agent Framework banner, and Semantic Kernel is shifting into security-and-deps mode — the kind of release pattern you see when a team is keeping production users safe while migration paths are being built elsewhere. Read in parallel with the eight-month silence at AutoGen, the picture is clear: Microsoft is collapsing three previous agent-framework efforts (SK, AutoGen, Semantic Workbench) toward one supported runtime.

Expect SK to stay on a security-and-deps cadence for at least another two quarters, with a hard deprecation timeline likely announced once Microsoft Agent Framework has feature parity. Anyone building net-new on Semantic Kernel today should plan a migration; existing deployments are safe for the moment but on borrowed roadmap time.