At the start of the 21st century, the field of condensed matter physics may seem to be quite well understood. From our fundamental knowledge we have been able to engineer all manner of materials. The question then may be ‘what now’? How deep really is our knowledge? Although we understand gases and crystals very well, liquids remain something of a puzzle. The strong interactions between constituent particles (like crystals), and their disorder (like gases) makes them something of a challenge. And as for ‘non-equilibrium’ phenomena, such as crystallization, melting and the glass transition, it is fair to say that we are only at the beginning. Why the gap in knowledge? The essential common issue is due to the limitations of visualising atoms. Atoms are simply too small for microscopy techniques, so we resort to ‘reciprocal space’ techniques. These reveal averaged information about an entire sample. Which is fine for crystals, where the periodic structure can be resolved beautifully. It is not fine for liquids, unless we assume that liquids are isotropic at all levels: it turns out they are not. And crystallisation, which starts, with small, nuclei of a few tens of atoms, is very difficult to access with scattering. So we would like to see atoms in ‘real space’, ie with a microscope. We use colloids, which, due to their Brownian motion, behave in the same was as atoms. Here is a picture of a crystal nucleating. We can resolve the colloids at the single particle level, track the coordinates with a computer, and obtain more detailed information than is possible with any other experimental technique. We have, for example, made real progress in understanding the structural mechanism of the glass transition and are currently working on crystallization mechanisms.