OpenAI describes its work to make ChatGPT better at programming languages with little training data. It has introduced new evaluations covering software engineering, SQLite and COBOL migration.
This article by Mallika Rao looks at the shift from microservices to agentic AI architectures. Where microservices split up functions, agent systems split up decisions, which brings new challenges for observability, reliability and design.
Chinese researchers studied whether chatbots answer differently depending on the language used. They report that prompts in Mandarin made the language models more open-minded.
To see what AI agents do when given no task, researchers repeatedly left copies of Claude Opus 4.7, nicknamed Eko, alone on a virtual desert island for 30 hours. What the agents did with their time looked a lot like play.
Meta engineers report better efficiency and accuracy in their data pipeline agents by keeping the orchestration layer deliberately thin. The orchestrator only routes work and enforces reliability checks, while specialized agents handle the domain reasoning.
Meta describes Generative Recommenders, which reason about ad content and user preferences together instead of scoring every ad, drawing on the world knowledge in LLMs. GenRecs tokenization is the data representation underpinning the approach.