Oracle Bans AI-Generated Code in OpenJDK While pgrust Achieves 300x Postgres Speedup
코딩/개발 | Sat Aug 08 2026 00:00:00 GMT+0000 (Coordinated Universal Time) | 2 sources
Oracle enforced a ban on AI-generated code contributions to OpenJDK, while pgrust delivered breakthrough Postgres performance through query engine optimization.
Analysis
[Oracle] enforced a ban on AI-generated code contributions to OpenJDK [1]
- Cited safety
- security
- and intellectual property risks as justification
- Prohibited submission of AI-generated code to project channels including repositories and pull requests
- Allowed personal LLM use for debugging and code review purposes
[Oracle Internal Policy] expanded internal AI code usage in contrast to external policy [1]
- Larry Ellison stated that AI models write Oracle's code
- Mike Sicilia mentioned the possibility of rapid development by smaller engineering teams using AI tools
- Investing $70 billion in data center expansion this year
[Oracle Credit Rating] downgraded to BBB- by S&P [1]
- Lowered to one level above junk rating
- Cited uncertain profitability against large-scale investment
[pgrust 0.2] achieved 300x acceleration over Postgres on analytical queries [2]
- 10x performance improvement over previous version
- 30% faster than Postgres on OLTP benchmarks
- Outperformed even Clickhouse on the Clickbench analytical benchmark
[pgrust Query Engine] redesigned the query engine based on batching, operator fusion, and SIMD [2]
- Query engine contributed approximately 10x of the total 300x performance improvement
- Minimized CPU and memory bandwidth usage
- Overcame Postgres's architecture designed in the 1980s disk I/O bottleneck era
[Postgres Performance Limitations] exposed unsuitability for modern hardware due to disk I/O era design [2]
- Many datasets now fit in RAM
- eliminating disk I/O bottlenecks
- NVMe improved disk speeds by hundreds of times over HDD
- Summing 500 million numbers takes Postgres 20 seconds vs Rust's 358ms (about 55x difference)