After more than 15 years of developments for astronomy, adaptive optics [AO] has become a powerful tools for ground based instrumentation at optical wavelengths. Installed in 2001 on one of the four 8-m telescopes of the ESO-VLT, NAOS is the paradigm of astronomical AO. This fully integrated and multi-purpose system coupled with the CONICA camera provides diffraction limited images in a large range of conditions (atmospheric turbulence and guide star flux) and applications. In the past five years, the system has generated numerous new and exciting astrophysical results among which the first direct image of an extrasolar planet. Despite of its first discovery, NAOS ultimate performance does not allow to properly cover this challenging area of current astronomy. Hence, new extremely high performance adaptive optics (XAO) system have to be designed to achieve the direct detection of extro-solar planets from the ground. Such systems will be coupled with coronographic and smart imaging technics. In the framework of the second generation instrumentation for the VLT, the "VLT-planet finder" [VLT-PF] instrument is currently designed and will be built within the next 4 years. The goal of this system is the direct imaging of hot Jupiter planets. Using a general framework, we propose here to depict the evolution of the AO systems from NAOS to VLT-PF. The performance and limitations of NAOS will be presented and discussed. Using the lesson learned on NAOS, new ideas and concepts are proposed for VLT-PF in order to overcome these limitations. The global philosophy of VLT-PF design will be presented and each new concept will be described. Using the same kind of approaches, we give some basic considerations on planet finder systems for the next generation of extremely large telescope (ELT).