Powerful radio galaxies — by displaying simultaneously active supermassive black hole (SMBH) and vigorous on-going star formation — are unique laboratories to study the impact of the SMBH feedback on its host galaxy. Their orientation in the plane of the sky allows the dusty torus to act as a natural coronograph, shielding us from the intense radiation from the accretion disc, thus, enabling the possibility to access simultaneously SMBH and host galaxy properties.
Using the efficient steep-spectrum selection technique from low-frequency radio surveys, we were able to build large samples of galaxies confirmed out to z=5.2 more than two decades ago. This provided us with the unique chance to study the physical properties of the progenitors of our local massive elliptical galaxies during the peak of the activity of the Universe (z~2-3). However, progress in finding more distant sources has been slow due to the lack of deep all-sky surveys at low frequencies but are crucial to understand how SMBHs grew so quickly in the young Universe.
We designed a new technique making use of the new generation of low-frequency radio surveys in order to push our samples beyond the twenty-years old limit of z=5.2. I will present the results of our multi-wavelength pilot programme making use of the VLT, ALMA and ATCA observations for a selection of sources from the GLEAM (an all-sky survey at 70-230MHz executed by the Murchison Wide-field Array). This technique has already shown its efficiency to confirm a z=5.55 source. Recent ALMA and JVLA observations place a second source of this sample at even higher redshift, potentially being the most distant radio galaxy detected so far, finally opening the opportunity to investigate radio galaxy evolution into the epoch of reionisation.