Core dump epidemiology: fixing an 18-year-old bug
The article describes how OpenAI identified two unrelated bugs causing crashes in their data infrastructure by analyzing core dumps and using epidemiological methods to trace the issues. The first bug involved hardware corruption on a specific Azure host, while the second was an 18-year-old race condition in GNU libunwind. By isolating the problems and eliminating false positives, the team resolved the issues through careful analysis and exclusion of the faulty host.
More from Openai
How RingCentral builds AI-native work from engineering to ops
See how RingCentral uses ChatGPT Work and Codex to accelerate AI product development and centralize operational intelligence across engineering and operations.
Daybreak models are now available on AWS
OpenAI and AWS are making Daybreak cybersecurity capabilities available through Amazon Bedrock to support enterprise security workflows.
Testing ads in ChatGPT
OpenAI begins testing ads in ChatGPT to support free access, with clear labeling, answer independence, strong privacy protections, and user control.
