Engineering at Meta
26 articles in the TruthFoundry index. Each links out to the original.
- Meta ports Android Cuttlefish virtual device framework to Apple Silicon Macs · Meta Reality Labs runs Android Cuttlefish natively on Apple Silicon macOS hardware.
- Intelligent Delivery System Enhances Ad Ranking Efficiency with Machine Learning · Intelligent Delivery optimizes ad ranking by using machine learning to allocate compute resources for better performance.
- Meta launches authenticated NTS network time service with open source code · Meta deployed RFC 8915 Network Time Security for its public internet time service.
- Meta launches F3 Lightning to accelerate ads ML feature production at scale · Meta introduced F3 Lightning, an infrastructure layer speeding ML feature releases 3.6x.
- Meta deploys production JiTTesting LLM system to catch pre-deployment code bugs · Meta has launched production JiTTests, an LLM-powered bug catching system for code changes.
- Meta launches next-gen Andromeda ads retrieval engine with GPU co-design · Meta's updated Andromeda system achieves 100x throughput lift for personalized ad retrieval.
- Meta's PrivacyLib: Enforcing Data Privacy at Scale · Meta's PrivacyLib is a common library that enforces data privacy across its systems by integrating with data access points.
- 2026 Python Typing Survey Reveals Shifts in Checker Use and AI Tool Integration · The 2026 Python Typing Survey found significant shifts in checker selection and increased AI tool use among developers.
- Meta unveils open AI networking innovations at 2026 OCP Global Summit · Meta presented open AI networking technologies at the 2026 OCP Global Summit.
- Meta launches GenRecs Tokenization for improved ad recommendation performance · Meta deployed GenRecs Tokenization delivering 0.35% conversion lift for advertisers.
- Meta's 'Thin Orchestrator' Pattern Improves AI Agent Reliability · Meta engineers improved AI agent accuracy by separating routing logic from domain reasoning in specialized subagents.
- Meta completes first phase of Prometheus gigawatt AI supercluster in Ohio · Meta has finished phase one of its Prometheus gigawatt-scale AI supercluster in New Albany, Ohio.
- Meta built Forum app using human-on-the-loop AI agent engineering · Meta developed its new Forum Groups app using agentic AI in engineering workflows.
- Meta infrastructure handles record 2026 World Cup global traffic surges · Meta platforms set all-time traffic records during the 2026 World Cup tournament.
- Sapienz research paper awarded ISSTA 2026 Most Influential Paper Award · 2016 Sapienz Android testing research receives 2026 ISSTA Impact Paper Award after industry adoption.
- Meta Unveils GEM 2026: Centralized AI Model Scales Ad Performance · Meta launches GEM 2026, a unified generative AI model optimizing ads across all platforms with 2.5x complexity.
- Meta Unveils Liquid-Cooled IcePack Racks and 102.4T IceCube Switches at OCP 2026 · Meta launches IcePack liquid-cooled racks and IceCube 102.4T switches to support gigawatt-scale AI clusters.
- Meta deploys Media over QUIC protocol for Instagram Live ingest · Meta has rolled out the open MoQ standard for Instagram Live video ingestion.
- Meta Deploys Elastic AI System to Turn Idle Capacity into Better Ads · Meta engineers built a system to use idle AI inference capacity for improved ad recommendations.
- Meta Launches Private Processing for AI Glasses to Ensure Data Privacy · Meta introduces Private Processing to keep AI glasses data secure in confidential cloud computing.
- Meta Open-Sources Rebalancer Library for Assignment Problems · Meta releases Rebalancer, a high-performance library for solving resource allocation and assignment problems.
- Meta Announces Petal: World's First Petabit-Class Transoceanic Subsea Cable · Meta partners with NEC and Sumitomo to build Petal, the first petabit-capable transoceanic subsea cable using multi-core fiber.
- Meta introduces ZGateway proxy layer for its ZippyDB key-value store · Meta deployed ZGateway proxy to manage traffic for its high-scale ZippyDB key-value store.
- Meta builds organizational AI second brain that preserves expert knowledge · Meta developed an AI agent that captures institutional expert knowledge without model retraining.
- Meta releases open MetaRoCE RDMA protocol for large-scale AI Ethernet networks · Meta launches open MetaRoCE RDMA transport built for million-GPU AI Ethernet clusters.
- Meta Unveils MTIA 300 Chip with Integrated NICs for Recommendation Models · Meta introduces MTIA 300, the first in-house training chip featuring built-in NICs to accelerate recommendation model training.