The Earth's climate system contains a highly complex interplay of numerous components, such as atmospheric greenhouse gases, ice sheets, and ocean circulation. Due to nonlinearities and feedbacks, changes to the system can result in rapid transitions to radically different climate states. In light of rising greenhouse gas levels there is an urgent need to better understand climate at such tipping points. Reconstructions of profound climate changes in the past provide crucial insight into our climate system and help to predict future changes. However, all proxies we use to reconstruct past climate depend on assumptions that are in addition increasingly uncertain back in time. A new kind of temperature proxy, the carbonate ‘clumped isotope’ thermometer, has great potential to overcome these obstacles. The proxy relies on thermodynamic principles, taking advantage of the temperature-dependence of the binding strength between different isotopes of carbon and oxygen, which makes it independent of other variables. Yet, widespread application of this technique in paleoceanography is currently prevented by the required large sample amounts, which are difficult to obtain from ocean sediments. If applied to the minute carbonate shells preserved in the sediments, this proxy would allow robust reconstructions of past temperatures in the surface and deep ocean, as well as global ice volume, far back in time. Here I propose to considerably decrease sample amount requirements of clumped isotope thermometry, building on recent successful modifications of the method and ideas for further analytical improvements. This will enable my group and me to thoroughly ground-truth the proxy for application in paleoceanography and for the first time apply it to aspects of past climate change across major climate transitions in the past, where clumped isotope thermometry can immediately contribute to solving long-standing first-order questions and allow for major progress in the field.