It is thought that the first generations of massive stars in the Universe were an important, and quite possibly dominant 1 , source of the ultra-violet radiation that reionized the hydrogen gas in the intergalactic medium (IGM); a state in which it has remained to the present day. Measurements of cosmic microwave background anisotropies suggest that this phase-change largely took place 2 in the redshift range z=10.8 ±1.4, while observations of quasars and Lyman-α galaxies have shown that the process was essentially completed 3,4,5 by z≈6. However, the detailed history of reionization, and characteristics of the stars and proto-galaxies that drove it, remain unknown. Further progress in understanding requires direct observations of the sources of ultra-violet radiation in the era of reionization, and mapping the evolution of the neutral hydrogen (H I) fraction through time. The detection of galaxies at such redshifts is highly challenging, due to their intrinsic faintness and high luminosity distance, whilst bright quasars appear to be rare It has long been recognised that GRBs have the potential to be powerful probes of the early universe. Known to be the end product of rare massive stars 11 , GRBs and their afterglows can briefly outshine any other source in the universe, and would be theoretically detectable to z ~ 20 and beyond 12,13 . Their association with individual stars means that they serve as a signpost of star formation, even if their host galaxies are too 5 faint to detect directly. Equally important, precise determination of the hydrogen Lyman-α absorption profile can provide a measure of the neutral fraction of the IGM at the location of the burst 9,10,14,15 . With multiple GRBs at z > 7, and hence lines of sight through the IGM, we could thus trace the process of reionization from its early stages.However, until now the highest redshift GRBs (at z = 6. Ground-based optical observations in the r, i and z filters starting within a few minutes of the burst revealed no counterpart at these wavelengths (see Supplementary Information (SI)).The United Kingdom Infrared Telescope (UKIRT) in Hawaii responded to an automated request, and began observations in the K-band 21 minutes post burst. These images ( Figure 1) revealed a point source at the reported X-ray position, which we concluded was likely to be the afterglow of the GRB. We also initiated further nearinfrared (NIR) observations using the Gemini-North 8-m telescope, which started 75 min after the burst, and showed that the counterpart was only visible in filters redder than about 1.2 µm. In this range the afterglow was relatively bright and exhibited a shallow spectral slope F ν ∝ ν -0.26 , in contrast to the deep limit on any flux in the Y filter (0.97-1.07 µm). Later observations from Chile using the MPI/ESO 2.2m telescope, Gemini South and the Very Large Telescope (VLT) confirmed this finding. The nondetection in the Y-band implies a power-law spectral slope between Y and J steeper than. This is impossible for dust at any redshift, and is a tex...