Alcides Fonseca, LASIGE integrated member, visited, on 13 February 2026, the Programming Methodology Group at ETH Zürich to present aeon, a programming language that uses liquid types, which prevent programmers from making a wide range of errors.

In his presentation, Alcides Fonseca showed how errors in the use of machine learning libraries, such as scikit-learn, can be avoided by using this library. Detecting errors in machine learning pipelines is very important because, unlike traditional software, programmes do not crash with incorrect inputs — they simply return an incorrect result or perform an action they should not. Detecting these errors earlier in the process leads to safer use of artificial intelligence, as well as saving money and time in software development.

In addition to machine learning, results were also presented on applying this technique to robotics, both with LiquidJava, a safer version of the Java language, and with ROSpec, a robotic software specification language that allows the identification of a very large set of ROS errors found on the ROSAnswers platform.