Minimal solvers are a fundamental building block of modern geometric vision pipelines, enabling efficient and robust estimation of camera motion and calibration parameters within RANSAC-based frameworks. Recent advances in monocular depth estimation have opened new opportunities for incorporating depth priors into classical geometric estimation problems. The central question is whether such depth information can be leveraged to improve camera pose estimation beyond what is achievable using image correspondences alone. In this talk, I will present our recent work on depth-aware minimal solvers for camera pose estimation. The main focus will be on RePoseD, a framework for relative pose estimation from image correspondences augmented with monocular depth predictions. Since modern depth estimators typically produce depths only up to unknown scale or affine transformations, we develop minimal solvers that jointly estimate camera motion together with the unknown depth parameters. The resulting methods apply to calibrated and uncalibrated camera settings and provide improvements in efficiency and robustness compared to existing depth-aware approaches. Extensive experiments demonstrate when and how monocular depth estimates can improve relative pose estimation in practice. Beyond relative pose estimation, I will briefly discuss other examples where approximate depth information simplifies geometric estimation, including camera pose estimation from object bounding boxes. I will conclude by arguing that geometry is not only a consumer of learned depth priors but can also serve as a tool for evaluating them. When depth predictions are used in downstream tasks such as pose estimation and structure-from-motion, the performance of the resulting geometric pipelines offers a natural benchmark that complements traditional pixel-wise depth evaluation and may better reflect practical utility. The talk highlights how imperfect but informative depth estimates can be integrated into minimal solvers, leading to faster, more robust, and more practical geometric estimation methods while also providing new perspectives on the evaluation of learned depth representations.