We constrain the parameters describing the kinematical state of the universe using a cosmographic approach, which is fundamental in that it requires a very minimal set of assumptions (namely to specify a metric) and does not rely on the dynamical equations for gravity. On the data side, we consider the most recent compilations of Supernovae and Gamma Ray Bursts catalogs. This allows to further extend the cosmographic fit up to $z=6.6$, i.e. up to redshift for which one could start to resolve the low $z$ degeneracy among competing cosmological models. In order to reliably control the cosmographic approach at high redshifts, we adopt the expansion in the improved parameter $y=z/(1+z)$ (as proposed in Class. Quant. Grav., 24 (2007) 5985). This series has the great advantage to hold also for $z>1$ and hence it is the appropriate tool for handling data including non-nearby distance indicators. We find that Gamma Ray Bursts, probing higher redshifts than Supernovae, have constraining power and do require (and statistically allow) a cosmographic expansion at higher order than Supernovae alone. Exploiting the set of data from Union and GRBs catalogues, we show for the first time a definitively negative deceleration parameter $q_0$ up to the $3 σ$ confidence level. We present also forecasts for realistic data sets that are likely to be obtained in the next few years.