From the archive
Monday, 2 November
Folded from 17 sources that day.
Sorted by quality and by how many sources covered it, the same for everyone.
-
Postgres 19 is late, and it's AI's fault
PostgreSQL 19 will likely slip to November after several large features were reverted late in beta. LLM-found bugs pushed back-patched commits up more than 60% year on year.
-
Rewarding defenders with bigger bug bounties
OpenAI is raising rewards in its Security Bug Bounty and adding bonuses for report quality, loyalty, and priority campaigns aimed at independent researchers.
-
Announcing the Special Interest Group (SIG) for Ray on Accelerators
Ray launches a vendor-neutral accelerator SIG for GPUs, TPUs, Trainium and other chips. Its initial roadmap covers testing infrastructure, hardware awareness and networking.
-
How Nvidia's New Nemotron Model Works
Breaks down Nvidia's Nemotron LLM, which combines Mamba and mixture-of-experts layers, multi-token prediction and storage reduction to cut compute and memory costs for agent workloads.
-
Import AI 479: Spacewar; low-latency tasks versus LLMs; Genie 4
Realtime-Bench is a 50-task benchmark for time-sensitive web and game tasks. The best model reached 36% under a real wall-clock latency, while paused-clock runs reached 70-80%.