Digital micromirror devices (DMDs) have found many scientific research applications. We present adaptive optics techniques exploiting the point spread function (PSF) of a DMD pixel to enhance the fidelity of image-projection-based laser machining. Femtosecond laser pulses with intensity profiles spatially shaped by a DMD were demagnified to a sample via a microscope objective, with ~10 DMD mirrors, each of width ~10µm, approximately projecting to the optical setup diffraction limit of ~1µm. A single DMD mirror then scales geometrically to dimensions well below the diffraction limit, permitting various techniques to enhance machining. By digitally shifting an intensity mask on the DMD between pulses while the sample remains static, machined features with resolutions below the single-exposure diffraction limit are produced (similar to pitch splitting multiple exposure techniques), with a reduction of <2.5x achieved in nickel. By combining digital image shifts with real-time sample image recognition algorithms, point-to-point positional accuracy is camera-resolution-limited (~500nm) rather than translation stage-limited. Furthermore, the PSF allows near-continuous intensity distributions rather than binary on/off intensity patterns, and have been used to produce variable-depth surface texturing (up to 40nm depth changes with 2µm period demonstrated in metals) features via single shots. Algorithms have been used to automate optical proximity corrections for arbitrary intensity masks in order to reduce machining errors due to optical filtering. These techniques are being combined to produce <1cm2 size, highly complex substrates for the production of biologically-friendly cell growth assays, with the viability of human bone stem cells on flexible substrates demonstrated.