Nearly 80% of harvested fish populations around the world are not assessed, and the situation is not much better in Australia, especially for recreational fisheries. This seems odd, considering that at least for coastal and freshwater fish species we have extensive and rapidly growing length observation data from visual and camera-based surveys, citizen science, recreational catches and, increasingly, AI-assisted monitoring. To turn these length-frequency observations into an evaluation of population status, we need to know how length distributions should look in the absence of fishing. In this talk I will present recent work demonstrating that: 1) average population- or species-level length distributions in the majority of fish can be approximated by a simple normal distribution; 2) the shape of this distribution is surprisingly similar across diverse species; and 3) in theory, unfished length distributions can be reconstructed from just a single measure of maximum length. I will also explain why maximum length (Lmax) estimates must account for sampling intensity, and introduce a new statistical method to estimate animal maximum lengths, implemented in the R package fishmax. Combined with the growing historical ecology research and rapid development of AI-based fish species and size estimation models, we should be entering a new era of size-based population assessment, and on track to have at least approximate population status estimates for most coastal and freshwater species.