Starting to teach inference early in an introductory statistics course means that the students have more time to assimilate the new concepts involved in estimation and hypothesis testing, especially if they are exposed to them in a sustained way throughout the semester. I teach a special section of the algebra-based course in which we start writing statistical hypotheses during the first week. We tell the students the general idea about testing hypotheses and that the details on how to calculate or approximate the p-value will depend on the context and the tools available. Randomization methods (permutation tests and bootstrapping) are introduced first because they require less background. After covering the basics of probability, the binomial distribution is used to do inference about one proportion. The classical methods using the normal, t-student and Chi-square distributions are studied at the end of the semester after these distributions have been introduced.