We’ve published a full post-mortem: https://
status.openai.com/incidents/ctrs
v3lwd797
… In short, a new telemetry service config flooded our Kubernetes control plane with requests, breaking DNS-based service discovery. We’ve made changes to prevent this happening again. Thank you for bearing with us!
SYSTEMS
-
OpenAI Post-Mortem: Telemetry Service Overloaded Kubernetes DNS
By
–
-
Custom Memory Systems Superior to General Approaches
By
–
Completely agree with this Bearish on general memory systems, need to be pretty custom
-
Lattice OS: AI system for real-time battlefield perception and autonomy
By
–
Lattice OS : Un système d’IA qui traite des données provenant de multiples capteurs, créant une vision en temps réel du champ de bataille et permettant aux systèmes autonomes de fonctionner harmonieusement. Les sous-traitants de défense traditionnels ne pouvaient pas rivaliser.
-
Linear Scaling AI Model Training Across Multiple Systems
By
–
After the single box ran, we scaled it to 2 and 16 systems with linear scaling. There were no model & code changes required.
Megatron
DeepSpeed
Sharding The model lives on a single block of memory. No need to break it up. The whole thing trains like one giant GPU. -
MemoryX Deploys 55TB DDR5 Memory in Commodity Server Format
By
–
MemoryX uses commodity DDR5 memory in a 1U server format, making it super easy to procure & configure. We employed one rack for CS-3 & networking and another rack for a 55 TB MemoryX partition.
-
Drones Shift Military Balance Toward Offensive Capabilities
By
–
Drones tip the balance of power towards offense, much like nuclear did. It may help to think of them as “autonomous bullets.”
-
Building Resilient and Diverse AI Solutions for Tomorrow
By
–
That’s the “Final Boss,” and why there needs to be a multiplicity of distinctly resilient options.
-
Character AI Tech Lead Discusses 30k QPS Scale Architecture
By
–
Tomorrow night I will be doing a fireside chat with James Groeneveld, a tech lead at Character AI Character AI does over 30k queries per second… one of the few GenAI apps to reach true scale Lots of interesting nuggets to dive into, RSVP! https://
lu.ma/rujejluk -
Fishbone Warehouse Layout Reduces Travel Time by 18%
By
–
Warehouse layout powered by 🐟 fishbone style = 18% less travel time
— Pascal Bornet (@pascal_bornet) 9 décembre 2024
❌ conventional warehouse: 20 positions in 48.5 seconds
✅ fish-bone warehouse: 20 positions in 40 seconds
Challenge: Designing unit-load storage areas (exclusive pallet-handling operations or bulk… pic.twitter.com/LHR64KFcqdWarehouse layout powered by fishbone style = 18% less travel time conventional warehouse: 20 positions in 48.5 seconds fish-bone warehouse: 20 positions in 40 seconds Challenge: Designing unit-load storage areas (exclusive pallet-handling operations or bulk
-
WordPress categories focused on AI topics
By
–
For anyone else reading along, tested this idea in a new post: