The oxygen abundance in the solar photosphere, and consequently the solar metallicity itself, is still a controversial question with far-reaching implications in many areas of astrophysics. This paper presents a new determination obtained by fitting the forbidden OI line at 6300 A with an *observational* 3D model. The approach presented here is novel because previous determinations were based either on 1D empirical stratifications or on 3D theoretical models. The resulting best-fit abundances are log E(O)=8.90 and log E(Ni)=6.15. Nevertheless, by introducing minor tweaks in the model and the procedure, it is possible to retrieve very different values, even down to log E(O)=8.70. This extreme sensitivity of the abundance to possible systematic effects is not specific to this particular work, but probably reflects the real uncertainty inherent to all abundance determinations based on a prescribed model atmosphere.