Europe 2026
The AI Skill I Rely On Daily — Priscila Andre de Oliveira, Sentry
Overview
This talk emphasizes that the most significant benefit of AI in large codebases is not code generation, but comprehension. The speaker shares personal experience using AI, particularly Claude, to navigate complex codebases at Sentry. The core argument is that AI acts as an invaluable teammate for understanding code, which is crucial for maintaining quality and avoiding technical debt, especially when working on projects that are critical to one's livelihood.
Who should watch
- AI engineers and developers working with large or unfamiliar codebases.
- Product Managers and engineers involved in shipping code who need to understand its intricacies.
- Individuals seeking to improve their productivity and understanding of complex systems using AI tools.
- Open-source maintainers and contributors who need to quickly grasp project structures and changes.
Key takeaways
- AI significantly enhances productivity by providing rapid answers to complex questions that would otherwise require extensive manual investigation or waiting for colleagues.
- The speaker's personal AI usage analysis revealed that 67% of prompts were for comprehension, while only 2% were for code generation, highlighting the importance of understanding over creation.
- A custom AI skill called "Catch Me Up" was developed to structure comprehension prompts into specific exploration modes like architecture, convention, feature trace, syntax, testing, and history.
- This skill helps in understanding new repositories, reviewing pull requests, and aligning mental models before prompting AI for tasks.
- Maintaining code quality is paramount, even when using AI; developers should avoid shipping subpar code, especially in projects that directly impact their income.
- AI can be viewed as a tireless teammate or the "cheapest senior engineer," encouraging users to ask questions freely to deepen their understanding.
- The complexity of Sentry's codebase, with hundreds of PRs daily and constant evolution, necessitates tools that facilitate rapid comprehension.
- AI-assisted comprehension is presented as a critical step before planning and implementation in the development process.
Notable quotes
*The biggest unlock from AI in a large code base isn't generation. It's comprehension.*
*AI is the teammate who never gets tired of your questions. So, there is no dumb questions.*
*Don't ship slop code into the code base that pays your salary. Ship keynote code.*
Unofficial community note. Prefer the recording for nuance.