The AI Native Dev - from Copilot today to AI Native Software Development tomorrow Transforming Debugging and Root Cause Detection with Sentry | David Cramer
29 snips
Aug 12, 2025 David Cramer, co-founder of Sentry, discusses innovations that help over 4 million developers ship software faster. He delves into the use of LLMs and how Sentry’s extensive app data network enhances debugging. David also touches on the emergence of tools like SEER for error detection and the challenges of managing complex data. He shares insights on modern coding practices, emphasizing the balance between automation and quality, while advocating for a proactive mindset to navigate evolving technologies.
AI Snips
Chapters
Transcript
Episode notes
Compress Data To Power LLM Diagnosis
- Sentry compresses and summarizes vast, structured app data and feeds those summaries into LLMs.
- That high-signal, interconnected dataset gives Sentry stronger diagnostic capabilities than generic log aggregators.
Embeddings Improved Deduplication Early
- David Cramer used embeddings to improve Sentry's fingerprinting and deduplication with large gains.
- He then built a trace-connected dataset and started SEER to enable LLM-driven root-cause analysis.
Don't Rely On Vibe Coding
- David Cramer warns that "vibe coding" is a myth and that LLM-driven work still requires specs, tests and code review.
- Use agents to accelerate boilerplate but keep humans in the loop for design and safety.