The supermassive black hole (BH) census remains very incomplete at the highest masses ($\gtrsim10^9$ $M_\sun$), limiting our understanding of BH/host galaxy co-evolution for the most massive galaxies. I present analysis of circumnuclear ionized atomic and both warm and cold molecular gas kinematics, as well as results of detailed gas-dynamical modeling, for a sample of nearby, luminous early-type galaxies (ETGs). Keck OSIRIS integral-field unit and \textit{Hubble Space Telescope} (\textit{HST}) STIS spectroscopy reveal H$_2$ 1$-$0 S(1) and H$\alpha$ in approximately Keplerian rotation within \mbox{NGC 1275}. We build and optimize models of dynamically-warm disk rotation to better constrain its BH mass $M_{\rm BH}=(1.20^{+0.32}_{-0.44})\times10^9$ $M_\sun$, although significant turbulent velocity dispersion complicates model fitting.

Dynamically cold molecular gas emission is free from nearly all of the systemics that plague $M_{\rm BH}$ determination when using dynamically warm tracers. I present Atacama Large Millimeter/submillimeter Array (ALMA) $\sim0\farcs3-$resolution CO(2$-$1) and continuum imaging of a sample of early-type galaxies (ETGs) that host circumnuclear dusty disks. Given their incredible data sensitivities, these dynamically cold disks yield the most sensitive possible tracers of the central gravitational potential. In several targets we detect central velocity upturns that suggest unresolved Keplerian rotation. We find these disks are formally stabilized against fragmentation and gravitational collapse. The continuum emission is in all cases dominated by a central, unresolved source and appears to be consistent with hot accretion.

I present follow-up, $0\farcs1-$resolution CO(2$-$1) imaging of one of these promising ETGs -- NGC 3258 -- to highly resolve its BH sphere of influence and map out the molecular kinematics in exquisite detail. Its very regular, nearly symmetric rotation enables the most detailed thin disk modeling to date. Moderate kinematic twists are incorporated using tilted-ring formalism, and the extended mass profile is amended to allow for a variable stellar mass density profile. Preliminary model optimization returns $M_{\rm BH}=2.23\times10^9$ $M_\sun$ with an anticipated total uncertainty of around 5\%.