<p>Manipulating cell identity through transcription factor-mediated reprogramming, induced pluripotency or directed differentiation holds promise for disease modelling and regenerative medicine. Yet the cells produced by these methods often do not fully recapitulate the molecular and functional characteristics of their native counterparts. Immaturity, low fidelity and heterogeneity remain barriers, limiting reliability for modelling human disease and therapeutic use. Recent advances in single-cell genomic technologies, integrative computational frameworks and emerging molecular recording tools are beginning to reveal the mechanisms underlying incomplete or inefficient reprogramming and highlight tractable failure points. Together, these approaches could support mechanism-guided protocol design and stepwise gains in fidelity, maturity and purity, potentially moving engineered cells towards clinical relevance and informing design principles for next-generation reprogramming strategies.</p>