Loading…
Meta
Meta Platforms is the parent corporation that owns major social and communication platforms including Facebook, Instagram, WhatsApp, and Threads.
- 22 posts summarised
- latest
The Good Stuff
Paid picks we use ourselves. Links carry a referral code — a subscription through one earns us a commission at no extra cost to you.
From their blog
Latest articles
How Meta Built Forum With AI-Native Engineering
Meta built Forum using AI-native engineering, treating an AI agent as a development participant rather than a faster typing tool. The team shifted from human-in-the-loop to human-on-the-loop, where the agent autonomously plans, implements, builds, tests, and corrects its work while the engineer sets goals and approves outcomes. Success was measured by verified, production-ready results, not code volume, with automated verification signals like build status and unit tests filtering output before review. Meta adapted the codebase for legibility, gave agents tools to inspect live view hierarchy directly instead of descriptions, and created a separate sandbox build to isolate prototype code from production. This approach helped Forum reach core feature parity with Facebook Groups in eight weeks and caught a memory issue during development that roughly halved memory use and page faults.
2026 World Cup: Breaking Records, Spiky Charts and Reliability at Scale
The 2026 FIFA World Cup produced the largest planned stress test of Meta’s infrastructure, with WhatsApp setting an all-time record of over 30 million messages per second during the final and delivering 8.4 billion messages in its peak five-minute window. Traffic analysis revealed that surges in sharing primarily occurred during breaks and final whistles, not during active gameplay, shaping capacity decisions. Instagram and Facebook saw live video watch time spike 90% and 242% respectively, while recommendation traffic rose 80% week-over-week. To absorb these spikes, Meta layered pre-provisioned capacity buffers, country-level graceful degradation, cascade prevention, and early warning signals across all serving tiers. Notably, the team did not shift load between regions or freeze deployments, relying instead on year-round reliability enhancements and phased runbook execution.
Sapienz Receives ISSTA 2026 Impact Paper Award
Sapienz, a multi-objective automated testing system for Android, received the ISSTA 2026 Most Influential Paper Award for its 2016 research by Ke Mao, Mark Harman, and Yue Jia. The system combines three goals: exploring more of an app, finding more crashes, and reducing the steps needed to trigger them, making failures easier to investigate. In an evaluation of 1,000 popular Google Play apps, Sapienz found 558 previously unknown crashes. Within 17 months of publication, it was deployed at Meta, where it now runs millions of tests daily, surfaces thousands of unique crashes each week, and helps safeguard hundreds of app releases each month. Sapienz also became the app-exploration engine for dynamic taint analysis at WhatsApp using synthetic conversation test accounts.
Meta’s Generative Ads Model (GEM): Scaling the Central Model and Compute Engine for Frontier Ads AI
Meta introduced GEM, the industry’s largest recommendation system foundation model, and its 2026 generation shifts from isolated per-domain optimizations on Facebook, Instagram, and Reels to a single central foundation model. The Kunlun architecture, co-designed with the latest GPU hardware, enables 2.5x model complexity and demonstrates LLM-like scaling laws with block-aligned sparse attention reducing sequence complexity. The LoopFM post-training technique transfers GEM’s internal representations to downstream models, doubling intelligence transfer at no additional compute cost or latency. GEM delivered 18.9% conversion improvement and 12% ad click-through rate improvement across all Meta ad surfaces and funnel stages, including 5.8% conversion uplift on Facebook, 4.1% on Instagram, and 1.6% on off-site conversion. Training efficiency doubled to 20-25% model FLOPS utilization while scaling total training FLOPs by 4x across thousands of GPUs using custom kernels and 5D parallelism.
OCP Summit 2026: Powering AI with Open Innovations in Scale-Up, Scale-Out, and Scale-Across Networks
At OCP Summit 2026, Meta unveiled next-generation open networking infrastructure for AI, including IcePack, its first liquid-cooled network rack, and IceCube, a 102.4T switch blade delivering 128×800G non-blocking capacity. The Flattened Planar Fabric (FPF) introduces a multiplane scale-out design where each GPU connects via eight independent 100G links to different switches, reducing latency and enabling larger clusters. Meta also shared its BAG super-spine for multi-site aggregation and Tahan, an Ethernet-based 51.2T scale-up switch for MTIA 300 systems. The company launched the Open Switch Platform Architecture (OSPA) initiative with Microsoft and Alibaba to standardize switch platform hardware and software interfaces. Additionally, Meta announced open-source implementations of MetaRoCE, its RDMA transport protocol optimized for AI cluster networks over Ethernet.
Shipping Media Over QUIC at Meta Scale for Instagram Live Ingest
Meta deployed Media over QUIC (MoQ), an emerging IETF standard built on QUIC, as the transport for live video ingest on Instagram Live across iOS and Android. MoQ provides independent streams that avoid head-of-line blocking, native relay-based fan-out and caching, tunable per-subscription latency from about 100 ms to several seconds, and track-aware delivery where relays can drop or prioritize objects. To replace the in-house RUSH protocol without disrupting production, Meta integrated the open-source moxygen library, placed a translation layer at edge endpoints to convert MoQ frames to RUSH, and used server-side routing for rapid experiment ramp-up and rollback. After months of experimentation, the MoQ path matched RUSH across all streaming-quality indicators and reliability measures, achieving metric parity as the rollout target. This production-tested foundation enables Meta to extend MoQ to end-to-end live delivery and real-time interactive AI services, simplifying multiple protocol stacks into one.
Turning Opportunistic AI Inference Capacity Into Better Ads Recommendations
Meta built a system that aggregates fragmented idle inference capacity from a globally constrained fleet into a shared opportunistic pool to enhance ads recommendations. Idle capacity arises from four mismatches: peak provisioning with demand variation, model development timing with capacity delivery, regional resource ratios, and accelerator types. Two elasticity levers exploit this preemptible, volatile capacity: Rank More Ads increases candidates evaluated by ranking stages, using both CPU and accelerator surplus, and Serve Better Models routes more requests to a more powerful model, using accelerator-only surplus. The system protects baseline serving through isolated routing, minute-scale feedback loops that adjust candidate volume and model coverage, and millisecond-scale value-aware load shedding that removes lower-value work during sudden preemption. This allows optional work to expand and contract with opportunistic capacity, driving better ad relevance and advertiser outcomes without requiring additional capacity orders.
NTS: Authenticated Time at Meta
Meta launched NTS (RFC 8915) authentication for its public time service at nts.meta.com, enabling clients to verify the origin and integrity of time packets. The service separates key establishment over TLS 1.3 from authenticated NTP, with cookie keys derived daily from a shared secret so servers hold no per-client state. Shortening TLS certificate lifetimes to 100 days in 2027 and 47 days in 2029 makes automated certificate renewal increasingly essential, creating a dangerous circular dependency on an authenticated clock. Meta open-sourced its NTS implementation and encourages mobile OS vendors to add client support, as Android and iOS currently sync time over unauthenticated UDP. The protocol does not prevent delay attacks, and device bootstrapping remains circular since NTS-KE requires TLS validation that itself demands a correct clock.
Bringing Private Processing to Meta AI Glasses
Meta AI glasses require cloud processing for advanced AI features like streaming transcription and long-term recall, but traditional cloud architectures expose data during use. Private Processing uses confidential virtual machines with Trusted Execution Environments (TEEs) on CPUs and GPUs to keep data encrypted even while being computed. Before sending context, glasses perform remote attestation, verifying the server’s TEE certificate against a public transparency ledger that records all deployed binary images. Non-targetable routing via blind-signed tokens and third-party OHTTP relays prevents tying requests to specific users, while encrypted storage ensures data is only accessible with user-provided keys. This infrastructure, already proven for WhatsApp and Meta AI, extends to glasses, with security audits by firms like NCC Group and a bug bounty program for independent research.
Open-Sourcing Rebalancer: A Generic, High-Performance Library for Solving Assignment Problems
Meta has open-sourced Rebalancer, a generic library for solving assignment problems that has been used internally for over nine years. Rebalancer separates problem specification from solution, using a three-step approach with modeling constructs like dimensions, partitions, scopes, and utilization, and a high-level spec API for common objectives and constraints. Problems are translated into an expression graph, and Rebalancer offers two solving techniques: an optimal solver using mixed integer programming (MIP) with commercial or open-source solvers, and a local search solver that evaluates millions of moves per second for large-scale problems. The local search solver handles problems with over a million objects and 5k bins, with an average solve time of 171 seconds. Rebalancer is used for roughly 40 million assignment problems daily across Meta’s infrastructure, and the open-source release includes a debugging UI called Rebalancer Explorer. The accompanying paper was published at OSDI’24.
Inside Petal: Building the World’s First Petabit-Class Transoceanic Subsea Cable
Meta announced Petal, the first transoceanic subsea cable to deliver petabit capacity, connecting France and the United States over 7,000 km and expected to enter service in 2029. It will be the first subsea cable system to deploy multi-core fiber technology at scale, using 2-core fiber in a 24 fiber-pair system to achieve 1 Pbps. The cable overcomes the Shannon Limit by adopting spatial division multiplexing with multi-core fiber, requiring ultra-pure synthetic silica for low attenuation and counter-propagation to minimize crosstalk. Petal’s repeaters use a single-body 96 amp design with a Fan-In/Fan-Out interface to transition between 2-core and single-core fiber for amplification, staying within existing 18 kV power feeding limits. Built with NEC, Sumitomo Electric, and Orange, Petal represents the largest generational capacity increase of any repeatered subsea system.
ZGateway: Learnings from Putting a Proxy in Front of ZippyDB
ZGateway is a stateless proxy tier introduced to unify traffic from ZippyDB’s large, diverse client fleet and address the reliability limits of direct access. By terminating TLS, authorizing requests, applying admission control and shaping, and routing through regional tiers, it replaces a many-to-many connection mesh with bounded client-to-gateway and gateway-to-server hops. Its embedded thick C++ client batches and coalesces shard requests, while a shared pipeline supports pure-proxy and read-through-cache tiers; per-use-case metrics, traces, and quota usage provide operational visibility. The gateway handles more than 1 billion operations per second, carries about 40% of ZippyDB traffic, and adds about 6% computational overhead to an average use case, with traffic projected to exceed 60%; production work also consolidated transaction state handling across 100% of transaction traffic without a reliability regression.
An Organizational Second Brain: Building an AI That Learns From Experts
Organizations often retain specialist knowledge in experts’ heads, leaving routine compliance assessments slow, inconsistent, and difficult to scale. The proposed “second brain” combines a structured, auditable knowledge architecture with a separate reasoning layer and a self-improvement loop that converts expert feedback into verified, regression-tested text updates without model retraining. High-density organizational positions, taxonomies, routing indexes, and gateway files are maintained in a dependency-linked filesystem, while sparse supporting material remains available through semantic or lexical retrieval. Composable recipes encode expert procedures, and an evaluation pipeline proposes auditable pull requests, retries failed fixes, and expands regression coverage after approval. After six weeks, SMEs reportedly found outputs useful almost always, assessments fell from days to minutes, and improvement cycles produced zero regressions; the architecture is presented as applicable beyond compliance.
MetaRoCE: A New RDMA Transport Built for AI-Scale Ethernet
Meta has designed MetaRoCE, a clean-sheet RDMA transport for AI workloads running on commodity Ethernet, targeting clusters that can reach million-GPU scale. Unlike standard RoCE, it treats packet loss and out-of-order delivery as normal: packets are sprayed across NIC-managed paths, written directly to their destinations, and retransmitted selectively without PFC or reorder buffers. It combines per-path telemetry and congestion windows with ECN-based sender control and receiver-driven fair-share hints, while a single connection supports multiple ordered streams and paths. On a 64-node AMD GPU cluster using RCCL collectives, Meta reports higher throughput and lower flow-completion times than RoCEv2, about 86% throughput at 1% packet loss, linear scaling across four- and eight-plane topologies, and autonomous recovery from plane failures. Meta is contributing the specification, reference implementation, and compliance suite to OCP, with a broader release planned for October at the 2026 OCP Global Summit.
MTIA 300: Meta’s First Training Chip with Built-in NICs and Communication-Offloading Engines
MTIA 300 is Meta’s first in-house training and inference accelerator optimized for ranking and recommendation models. These workloads require frequent AllReduce, AllToAll, and AllGather operations because embedding tables can contain over 99% of model parameters, making communication a bottleneck that competes with computation on GPUs. Its package-integrated network uses two chiplets with twelve custom 800 Gbps RDMA NICs, while 16 message engines and near-memory compute handle collectives independently of the 12×6 processing-element grid. Co-designed HCCL compiles collectives into dependency-aware subgraphs for autonomous execution, with topology-aware algorithms spanning scale-up and scale-out links. In production, HCCL reaches up to 940 GB/s within a rack, and a 150-billion-parameter model across 40 accelerators completed communication 3.9 times faster than an equivalent GPU cluster; concurrent GEMMs degraded less than 0.5%.
How We’re Building Scam Alert on WhatsApp With End-to-End Encryption and Verifiability Guarantees
WhatsApp presents an early technical overview of Scam Alert, an optional feature that uses an on-device machine learning model to identify potential scam messages from non-contacts while keeping message content on the device. The warning appears only to the user, who can block, report, continue, or mark a chat trusted; reporting and sharing require explicit user action. To measure performance, the system sends locally aggregated warning and user action counts through confidential federated analytics using Trusted Execution Environments and differential privacy, rather than transmitting raw signals. Model versions are recorded on a public transparency ledger, while published weights, client-side logs, and an expanded Bug Bounty program support independent inspection of delivery and behavior.
From User Sequences to Scaling Laws: A Multi-Stage Architecture for Meta’s Ads Ranking
Meta’s ads ranking systems process millions of candidates per second under millisecond latency budgets, creating tension between richer sequence models and efficient serving. The post presents a multi-stage architecture that moves long-history user modeling into asynchronous offline transformers, caches user-level embeddings, and combines them with fresh user and ad signals in an online ranking model. Dense tokenization places sparse features and behavioral sequences in one vocabulary, while target-aware multi-head attention learns higher-order interactions between historical behavior and the candidate ad. On real-world ads traffic, performance follows a log-linear relationship with compute across depth, enrichment, width, and sequence length, with balanced growth and diverse sequences identified as scaling levers. Together with broader model innovations, the approach contributed to a cumulative 6% conversion lift on Instagram, 3% on Facebook, and 3.5% lift in Facebook ad clicks, and serves as a core component of GEM.
GEM Training: How Meta Doubled the Efficiency of Its LLM-Scale Ads Foundation Model
Meta’s Generative Ads Recommendation Model (GEM), a hybrid foundation model for ads recommendations on Instagram and Facebook, trains across several thousand latest-generation GPUs and combines trillions of sparse embedding parameters with billions of dense parameters. The post frames its engineering problem around recommendation-specific jagged and asymmetric sequences, memory-bound operations, numerical sensitivity, communication overhead, activation pressure, and load skew that make conventional LLM training techniques insufficient. Its co-design approach combines custom kernels such as Jagged Flash Attention (JFA), BlockAttention, and Generalized Dot-Product Attention (GDPA), mixed ultra-low-precision training including MXFP8 attention and MLP, topology-aware 5D parallelism, SM-free collectives, and memory and load-balancing techniques. The reported result is a doubling of end-to-end training efficiency to 20–25% Model FLOPs Utilization (MFU) while total training FLOPs increased fourfold over 12 months; Base Batch Shuffling (BBS) alone delivered a reported 4% efficiency gain, with 4% QPS improvement and 4% peak memory reduction.
Exploring Hierarchical Interest Representation For Meta Ads Deep Funnel Optimization
Hierarchical Interest Representation is an upstream representation layer for Meta Ads that addresses sparse deep-funnel feedback by connecting users’ inferred interests with advertisers, products, services, and other ads entities. It models a typed, heterogeneous engagement graph with a transformer architecture, sparse attention for long-range relationships, bias-aware attention, self-supervised cross-view distillation, and multimodal content features processed through vision or language models. The system projects the raw graph into learned interest primitives and cascading super-graphs, producing universal embeddings and Bag-of-Meaning tokens at multiple granularities. Temporal cutoffs and chronological splits prevent future-information leakage, while a pipelined online graph engine reportedly achieved a 30x wall-clock speedup over a synchronous baseline during testing. Trained end-to-end on billions of interactions, the representations are designed for personalization, retrieval, supervision, and ranking integrations including GEM, Andromeda, and the Adaptive Ranking Model.
Modernizing the Meta Ads Service With an Open-Source Kernel Scheduler
Meta’s ads serving fleet processes more than 5 million requests per second, so millisecond-scale latency changes can affect ads relevance and advertiser ROI. During a move from kernel 6.4 with CFS to kernel 6.9, the team observed a latency regression associated with EEVDF and used sched_ext, an upstream BPF-based framework, to encode ads-specific scheduling priorities. The policy soft-partitions CPUs between latency-critical request-path threads and less-sensitive work, adjusts pool sizes with load-based heuristics, and runs as a user-space binary that loads BPF, enabling policy changes without rebuilding or reinstalling the kernel. Initial results were a 28% reduction in ads retrieval p99 latency, 3.28 megawatts of fleet power savings, and a 1.1% increase in weighted ads ranked; two later user-space updates delivered an additional 60% p99 reduction and an 18% decrease in critical-path timeout errors, with iterations shipping in days rather than months.
Topics they write about
All topics →Weekly digest
The best engineering reads, once a week.