Local-type primordial non-Gaussianity remains one of the cleanest windows onto inflation. Spectroscopic 3D surveys offer large volumes and many modes, but are hampered on the largest scales by cosmic variance, uncertainty in the bias and shotnoise. Zero-bias tracers – tracers with effective linear bias approaching zero – promise a remedy: by reducing sampling variance these tracers make the scale-dependent signature from f_NL stand out more cleanly. Recent work (using the AbacusPNG suite) shows that the inflation-specific bias parameter, b_\phi, which quantifies how tracers respond to long-wavelength potential fluctuations, can be modelled much more tightly by including secondary halo/galaxy properties (concentration, environment, non-local bias). This reduces the uncertainty in b_\phi (and hence in f_NL) by factors of 2–5 relative to cruder assumptions.Our ongoing project aims to apply these ideas to DESI and SPHEREx mocks, hunting for zero-bias or near zero-bias tracer populations that still retain enough number density and redshift fidelity to be useful. I will finally comment on a new emerging measurement that has the power to transform searches for f_NL: cosmic velocity measurements via the CMB secondary, kSZ. By combining density and velocity tracers, or making use of velocity directly, we can hope to further reduce uncertainties on f_NL and bring us close to the coveted unity precision, which has the potential to question our current paradigm or else rule out wide classes of inflationary models.