In order to explain rapid light curve variability without invoking a variable source, several authors have proposed "minijets" that move relativistically relative to the main flow of the jet. Here we consider the possibility that these minijets, instead of being isotropically distributed in the comoving frame of the jet, form primarily perpendicular to the direction of the flow, as the jet dissipates its energy at a large emission radius. This yields two robust features. First, the emission is significantly delayed compared with the isotropic case. This delay allows for the peak of the afterglow emission to appear while the source is still active, in contrast to the simplest isotropic model. Secondly, the flux decline after the source turns off is steeper than the isotropic case. We find that these two features are realized in gamma-ray bursts (GRBs): 1. The peak of most GeV light curves (ascribed to the external shock) appears during the prompt emission phase. 2. Many X-ray light curves exhibit a period of steep decay, which is faster than that predicted by the standard isotropic case. The gamma-ray generation mechanism in GRBs, and possibly in other relativistic flows, may therefore be anisotropic.

The mechanism behind the launching of gamma-ray-burst (GRB) jets remains debated resulting in large uncertainty over the jet composition. Both magnetohydrodynamical and neutrino annihilation models have been proposed for the energy extraction in a black hole/accretion-disc central engine. In particular, for the extreme accretion rates $\dot M\sim 0.1-1$~ M$_\odot$s$^{-1}$ expected for bursts of duration $T\lesssim 100$~s, the disc can be an efficient neutrino emitter. Neutrino-antineutrino annihilation results in an energy deposition rate at the jet that can, in principle, account for the burst's energetics. Recent discoveries of X-ray flares hours after the burst and of ultra-long GRBs suggest that GRB activity can last for $\sim 10^4$~s or longer. These long-lived events have fluence similar to that of classical GRBs. In view of these findings, we re-evaluate the neutrino annihilation model. We derive the maximum possible energy of a neutrino-powered jet as a function of the burst duration and show that the available energy drops fast for longer bursts. For a standard choice of the parameters, the model falls short by three to four orders of magnitude in explaining the observed energetics of events that last longer than $\sim 10^3$~s.