From autocomplete to conversational agents, AI is redefining how we write, explore, and understand code. But what happens if we give an LLM full access to our interactive runtime system?

In this talk, I’ll present SemanticText, a new framework that integrates large language models (LLMs) into Squeak/Smalltalk. SemanticText provides the foundation for building and prototyping conversational agents, semantic search tools, and retrieval-augmented generation workflows, while trying to align with existing Smalltalk values such as liveness and explorability.

Building on this foundation, I’ll showcase different ideas for improving the programming experience of offered by Squeak. I will present experimental integrations of LLMs into Squeak’s existing tool set for interacting with code, browsing documentation, and navigating the Squeak developers’ mailing list. Moving further, I will introduce our concept of a semantic exploratory programming system, which aims to streamline and augment the workflow of exploratory programmers, and we will talk to objects in natural language to explore and debug systems with the help of AI.

After the talk, there will be time for any questions and joint experiments.

---

Christoph Thiede is a member of HPI’s Software Architecture Group with a strong interest in developer productivity and tools. As a core developer of Squeak/Smalltalk, he enjoys delving deep into its ecosystem and contributing to everything from its simulation machinery to libraries and debugging tools. His master’s thesis investigates novel approaches to leveraging generative AI for augmenting exploratory programming.

---

Meeting link: https://www.meetup.com/ukstug/events/302453224