Abstract：Determining the cosmic evolution of cold molecular gas in galaxies is fundamental for our understanding of galaxy evolution. However, obtaining such measurements for large samples of galaxies at z~1--6 remains challenging even in the era of ALMA. Recent studies of the cosmic gas evolution focus on redshifts below z=3 reporting conflicting results, and cosmological simulations still struggle to fit the observations. As (sub-)millimeter dust continuum has now been established as a reliable tracer of the cold gas, and it can be much more efficiently observed than line tracers (e.g. CO), we have started a concerted effort to automatically mine the public ALMA archive for the COSMOS field (A3COSMOS). A3COSMOS provides (sub-)millimeter detections and subsequent galaxy properties in a systematic way that allows us to quantify systematic biases in using dust continuum. Our database currently includes about 2000 ALMA images covering 280 sq. arcmin primary beam area, with over 1500 ALMA detections based on well-characterized statistics. Using rich ancillary data available we identify ~1000 galaxies with bona-fide properties lying at z~1--6. We combine this unique catalog of cold gas measurements with more than 500 local (z~0) and high-redshift (z<3) galaxies with CO and/or dust continuum detections in the literature to explore the correlations between gas mass fraction, molecular gas depletion time, redshift, stellar mass and star formation rate. Our analysis reveals a linear evolution of the gas fraction with cosmic time, in line with the evolution of galaxies’ “star-forming main sequence”. This evolution further implies a cosmic molecular gas density evolution consistent with previous observational works at z<1-3 and the predictions of cosmological simulations even at z>5--6. As ALMA accumulates more observations in the future, the cold molecular gas evolution at z>5--6 will eventually be constrained by A3COSMOS.