There is no doubt that AI based tooling can increase developer productivity. But (as ever) the devil is in the details. Do these gains come from code generation, refactoring, or optimisation? What about translating between languages and language versions? Are applications of LLMs hampered by the aging parametric knowledge base of these models? If so can this knowledge be updated, given ethical and ecological constraints? This talk will investigate these issues, using the JetBrains AI Assistant as an exemplar. We will combine practical coding demos with architectural insight to explore the impact of AI on the industry today, and its viability into the future.