AI Debugging: Beyond Bug Fixing to Causal Understanding - Episode Hero Image

AI Debugging: Beyond Bug Fixing to Causal Understanding

Original Title:

Resources

Resources & Recommendations

Books

  • "Augmenting Debugging with Large Language Models" by University of Massachusetts Amherst, AWS, and Williams College - This research paper describes ChatDBG, a tool that hooks up conventional software debuggers to LLMs, allowing users to ask natural language questions about program behavior.

Research & Studies

  • "Debug Like a Human: A Large Language Model Debugger via Verifying Runtime Execution Step by Step" (University of California San Diego) - This paper introduces LDB (LLM Debugger), which demonstrates that giving a coding agent the ability to step through code and verify its dynamic behavior can make it smarter and reduce errors.

Tools & Software

  • s-trace - A system-level utility mentioned as a tool that can be used in debugging.
  • gdb - A common debugger used to step through code and inspect variables.
  • Valgrind - A memory debugging and profiling tool.
  • perf - A performance analysis tool for Linux.
  • ChatDBG - A research-based tool that integrates LLMs with software debuggers, enabling natural language queries about program execution.
  • LDB (LLM Debugger) - A tool that allows LLMs to step through and verify code execution, enhancing their debugging capabilities.

Organizations & Institutions

  • IEEE Computer Society - The organization that sponsors Software Engineering Radio.
  • IEEE Software Magazine - A publication from IEEE that co-sponsors Software Engineering Radio.
  • University of Massachusetts Amherst - Involved in the research and development of ChatDBG.
  • AWS - Involved in the research and development of ChatDBG.
  • Williams College - Involved in the research and development of ChatDBG.
  • University of California San Diego - Involved in the research and development of LDB.
  • Undo - Mark Williamson's company, specializing in kernel-level low-level Linux embedded development and time-travel debugging.
  • Anthropic - Developed the Model Context Protocol (MCP).

Websites & Online Resources

  • se-radio.net - The official website for the Software Engineering Radio podcast.
  • computer.org/software - The online presence for IEEE Software Magazine.
  • github - Mentioned as a common platform for issue tracking.
  • Stack Overflow - A popular Q&A site for programmers, mentioned as a previous source for problem-solving before LLMs.
  • seradio.slack.com - The Slack channel for SE Radio listeners.

Other Resources

  • OpenTelemetry - A set of APIs, SDKs, and tools used to instrument, generate, collect, and export telemetry data (metrics, logs, and traces) for modern applications.
  • Model Context Protocol (MCP) - A plugin architecture that allows LLMs to interact with external tools, initially developed by Anthropic.

---
Handpicked links, AI-assisted summaries. Human judgment, machine efficiency.
This content is a personally curated review and synopsis derived from the original podcast episode.