Modern developers struggle with notification overload from various platforms, leading to "triage fatigue." Existing AI tools often worsen this by generating more text. CalmQueue offers a radical alternative, focusing on showing less.It's an open-source, local-first notification assistant built with a "Slow and Steady" ethos. CalmQueue ingests diverse notification streams and groups repetitive, automated, or duplicate alerts using semantic similarity. From these, it synthesizes exactly three actionable priorities (P0, P1, P2) for the workday.Every priority is grounded with verifiable verbatim quotes from the original source messages. Crucially, CalmQueue enforces strict human-in-the-loop oversight; draft responses require a one-click human review. This prevents silent external mutations and ensures user control.The system significantly reduces the displayed volume of notifications, offering a quiet, high-contrast dashboard. An "Evidence Grounding Drawer" provides source details, and a "Review & Approve Modal" facilitates human review of draft responses. CalmQueue is 100% open-source under the MIT license, with a live interactive web demo available.Its architecture includes a semantic cluster engine, a task reasoning and extraction engine, and an evidence grounding and ranker. The design prioritizes confidentiality, data sovereignty, and offline resilience by using local open-weight LLMs like Ollama. This approach eliminates reliance on costly cloud APIs and ensures auditable, verifiable AI behavior.Benchmarks show significant displayed volume reduction and low processing latency, enabling users to efficiently triage their backlog. CalmQueue embodies the "Slow and Steady" theme by promoting cognitive subtraction and safeguarding user attention.