# Engineering at Meta

Engineering at Meta Blog

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## How Meta Uses AI Agents to Scale ML Innovations Across Ads Ranking Models

DevFeed: [How Meta Uses AI Agents to Scale ML Innovations Across Ads Ranking Models](<https://devfeed.tech/articles/how-meta-uses-ai-agents-to-scale-ml-innovations-across-ads-ranking-models-141155.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/11/18/ml-applications/ai-agents-to-scale-ml-innovations-across-ads-ranking-models/>)

Author: Rahul Suresh; Aayush Aggarwal; Aayush Prakash; Peggy Yao; Shiblee Sadik; Djordje Gligorijevic; Akshay Pardeshi; Yao Xuan; Anirudh Agrawal; Matan Levi

Published: 2026-11-18T17:00:00Z

Content type: article

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>), [Meta ML Applications](<https://devfeed.tech/sources/meta-ml-applications.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [MLOps](<https://devfeed.tech/topics/mlops.md>), [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>)

Tags: [ads](<https://devfeed.tech/tags/ads.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [designs](<https://devfeed.tech/tags/designs.md>), [devinfra](<https://devfeed.tech/tags/devinfra.md>), [efficiently](<https://devfeed.tech/tags/efficiently.md>), [engineer](<https://devfeed.tech/tags/engineer.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [first](<https://devfeed.tech/tags/first.md>), [fleet](<https://devfeed.tech/tags/fleet.md>), [meta](<https://devfeed.tech/tags/meta.md>), [ml](<https://devfeed.tech/tags/ml.md>), [ml-applications](<https://devfeed.tech/tags/ml-applications.md>), [models](<https://devfeed.tech/tags/models.md>), [ranking](<https://devfeed.tech/tags/ranking.md>), [ranking-models](<https://devfeed.tech/tags/ranking-models.md>), [scale](<https://devfeed.tech/tags/scale.md>), [uses](<https://devfeed.tech/tags/uses.md>)

### AI overview

Meta describes using AI agents to spread proven ads ranking techniques across a diverse model fleet, where adaptations often require model-specific changes and training runs. The workflow has three stages: discovery prioritizes technique-model candidates; generalization adapts techniques through shared templates or bespoke work on individual models; adoption runs and monitors experiments, diagnoses failures, and allocates compute among concurrent jobs. Agents combine LLM reasoning with ML tools and a shared record of prior runs, while engineers review decisions and approve what ships. Meta reports agents handle over 60% of propagation; early results include 3x greater engineering productivity, a median 4x reduction in training compute for select combinations, and estimated compute savings above $45M by year-end. Techniques must improve offline predictive metrics such as normalized entropy without breaching cost guardrails, then show a statistically significant online A/B win. Meta aims to cut propagation time from over a year to three months; these are reported results and goals, not fleet-wide guarantees.

### Source excerpt

Our previous post in this series introduced the Ranking Engineer Agent (REA), an autonomous system that generates hypotheses, designs experiments, and proves new machine learning (ML) techniques on ads ranking models. Now, we're sharing how we propagate ranking model improvements like architectural advancements and training optimizations efficiently across a large and diverse fleet of ads [...] Read More... The post How Meta Uses AI Agents to Scale ML Innovations Across Ads Ranking Models appeared first on Engineering at Meta.

## Design Once, Deploy at Scale: Building a Template-Driven ML Ecosystem

DevFeed: [Design Once, Deploy at Scale: Building a Template-Driven ML Ecosystem](<https://devfeed.tech/articles/design-once-deploy-at-scale-building-a-template-driven-ml-ecosystem-141156.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/11/18/ml-applications/design-once-deploy-at-scale-template-driven-ml-ecosystem/>)

Author: Anirudh Agrawal; Djordje Gligorijevic; Yao Xuan; Peggy Yao; Jiang Liu; John Martabano Landy; Rahul Suresh

Published: 2026-11-18T16:59:38Z

Content type: article

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>), [Meta ML Applications](<https://devfeed.tech/sources/meta-ml-applications.md>)

Topics: [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [model architecture](<https://devfeed.tech/topics/model-architecture.md>), [Template](<https://devfeed.tech/topics/template.md>)

Tags: [all](<https://devfeed.tech/tags/all.md>), [approach](<https://devfeed.tech/tags/approach.md>), [bottleneck](<https://devfeed.tech/tags/bottleneck.md>), [building](<https://devfeed.tech/tags/building.md>), [but](<https://devfeed.tech/tags/but.md>), [deploy](<https://devfeed.tech/tags/deploy.md>), [design](<https://devfeed.tech/tags/design.md>), [devinfra](<https://devfeed.tech/tags/devinfra.md>), [ecosystem](<https://devfeed.tech/tags/ecosystem.md>), [ecosystems](<https://devfeed.tech/tags/ecosystems.md>), [efficiently](<https://devfeed.tech/tags/efficiently.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [inventing](<https://devfeed.tech/tags/inventing.md>), [large](<https://devfeed.tech/tags/large.md>), [meta](<https://devfeed.tech/tags/meta.md>), [ml](<https://devfeed.tech/tags/ml.md>), [ml-applications](<https://devfeed.tech/tags/ml-applications.md>), [model](<https://devfeed.tech/tags/model.md>), [models](<https://devfeed.tech/tags/models.md>), [more](<https://devfeed.tech/tags/more.md>), [scale](<https://devfeed.tech/tags/scale.md>), [template](<https://devfeed.tech/tags/template.md>)

### AI overview

Meta describes Standard Model Templates (SMTs) as a way to reduce fragmentation across its ads ranking and retrieval models: shared components define reusable model patterns, so teams can develop techniques at the template level and propagate them to groups of models. Multi-model optimization evaluates candidate techniques on representative models for aggregate gains and regression rates, then checks transfer on held-out models before integration. Meta reports that over two years the approach reduced per-model engineering time by 92%, shipped more than 5,000 technique-model combinations and 80 techniques, improved offline quality by 0.63% on average at neutral serving capacity, and delivered 0.86% cumulative online lift. The text says templates also provide a structured starting point for agents, while noting that coverage, reliability, explainability, and preserving useful flexibility remain ongoing challenges.

### Source excerpt

As ML ecosystems grow to a large number of production models, the bottleneck is no longer inventing new techniques to improve these models, but rather shipping them reliably and efficiently across a fragmented model space, creating a rising tide that lifts all the boats. At Meta, we built a template-driven approach that standardizes model design [...] Read More... The post Design Once, Deploy at Scale: Building a Template-Driven ML Ecosystem appeared first on Engineering at Meta.

## Cuttlefish on Mac: Running Android's Virtual Device Framework Natively on Apple Silicon

DevFeed: [Cuttlefish on Mac: Running Android's Virtual Device Framework Natively on Apple Silicon](<https://devfeed.tech/articles/cuttlefish-on-mac-running-android-s-virtual-device-framework-natively-on-apple-silicon-134031.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/11/13/android/cuttlefish-on-mac/>)

Author: Hadrien Zalek; Joe Knox; Alex Chong; Marcel Apfelbaum

Published: 2026-11-13T17:00:39Z

Content type: article

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>)

Topics: [macOS](<https://devfeed.tech/topics/macos.md>), [virtualization](<https://devfeed.tech/topics/virtualization.md>), [virtual machines](<https://devfeed.tech/topics/virtual-machines.md>), [Android](<https://devfeed.tech/topics/android.md>), [aosp](<https://devfeed.tech/topics/aosp.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [android](<https://devfeed.tech/tags/android.md>), [apple](<https://devfeed.tech/tags/apple.md>), [apple-silicon](<https://devfeed.tech/tags/apple-silicon.md>), [backend](<https://devfeed.tech/tags/backend.md>), [cuttlefish](<https://devfeed.tech/tags/cuttlefish.md>), [device](<https://devfeed.tech/tags/device.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [framework](<https://devfeed.tech/tags/framework.md>), [labs](<https://devfeed.tech/tags/labs.md>), [mac](<https://devfeed.tech/tags/mac.md>), [macos](<https://devfeed.tech/tags/macos.md>), [meta](<https://devfeed.tech/tags/meta.md>), [qemu](<https://devfeed.tech/tags/qemu.md>), [reality-labs](<https://devfeed.tech/tags/reality-labs.md>), [virtual](<https://devfeed.tech/tags/virtual.md>)

### AI overview

Meta ported Android's Cuttlefish virtual-device framework to run natively on Apple silicon Macs, enabling developers to use the same higher-fidelity virtual devices in local development and CI. The port revived Cuttlefish's QEMU backend and addressed three Linux-specific gaps: vhost-user components were made buildable on macOS, QEMU's Hypervisor.framework backend gained more memory-region tracking, and native display, build, and signing support were added. The Cocoa display supports BGRA rendering, pointer-to-touch input, and configurable geometry. Mac builds require a JIT entitlement, code signing, and notarization. The unified platform reduces the burden of maintaining Goldfish and Cuttlefish separately and supports capabilities including Bluetooth, Wi-Fi, cellular, 3D graphics, OTA testing, fastboot, and snapshots. Meta shares the work through its qemu-wearables fork and cuttlefish-wearables patch set, while exploring upstream contributions.

### Source excerpt

How Meta's Reality Labs brought Cuttlefish and its QEMU backend to macOS and Apple silicon. [...] Read More... The post Cuttlefish on Mac: Running Android's Virtual Device Framework Natively on Apple Silicon appeared first on Engineering at Meta.

## Intelligent Delivery: Making Every Unit of Compute Serve Better Ads

DevFeed: [Intelligent Delivery: Making Every Unit of Compute Serve Better Ads](<https://devfeed.tech/articles/intelligent-delivery-making-every-unit-of-compute-serve-better-ads-129628.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/11/11/ml-applications/intelligent-delivery-serve-better-ads/>)

Author: Zhaojun Zhang; Qinwei Zhu; Gang Li; Jianwei Xiao; Yanan Wang; Hongmei Li; Deepak Vijaywargi; Ke Pan; Li Yu

Published: 2026-11-11T17:00:38Z

Content type: article

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>), [Meta ML Applications](<https://devfeed.tech/sources/meta-ml-applications.md>)

Topics: [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [model-serving](<https://devfeed.tech/topics/model-serving.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [A/B Testing](<https://devfeed.tech/topics/a-b-testing.md>)

Tags: [accelerators](<https://devfeed.tech/tags/accelerators.md>), [adaptive](<https://devfeed.tech/tags/adaptive.md>), [ads](<https://devfeed.tech/tags/ads.md>), [advanced](<https://devfeed.tech/tags/advanced.md>), [agentic](<https://devfeed.tech/tags/agentic.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [amd](<https://devfeed.tech/tags/amd.md>), [compute](<https://devfeed.tech/tags/compute.md>), [delivery](<https://devfeed.tech/tags/delivery.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [impact](<https://devfeed.tech/tags/impact.md>), [llm](<https://devfeed.tech/tags/llm.md>), [meta](<https://devfeed.tech/tags/meta.md>), [ml-applications](<https://devfeed.tech/tags/ml-applications.md>), [models](<https://devfeed.tech/tags/models.md>), [ranking](<https://devfeed.tech/tags/ranking.md>), [ranking-models](<https://devfeed.tech/tags/ranking-models.md>), [system](<https://devfeed.tech/tags/system.md>)

### AI overview

Meta's Intelligent Delivery allocates finite serving compute toward the ad requests, ranking stages, regions, and hardware where it is expected to have the greatest marginal value. It reports about a 10% additional ads ranking performance improvement since 2017, alongside model and system advances. The system uses shared deep learning models to predict request outcomes, compute use, and latency, then applies layered controls: shifting traffic between regions, matching workloads to measured hardware efficiency, and trimming lower impact work when host latency budgets tighten. These coupled decisions are validated in end to end hardware tests. AI agents now monitor system changes, draw on code, design documents, and experiment history, and propose and run optimization experiments; engineers review and approve changes before release. Agents contributed to about 20% of optimizations in the first half of 2026, and the post says this reduced the optimization cycle from months to weeks. Joint real time optimization across all stages and adapting allocation for GPU based ranking models remain open work.

### Source excerpt

Ads ranking models are growing toward LLM scale faster than the hardware to run them can keep up. An extra unit of compute has more impact on some ad requests, models, and chips than on others. The challenge is deciding where to allocate it to get the best outcome for the entire system. We built [...] Read More... The post Intelligent Delivery: Making Every Unit of Compute Serve Better Ads appeared first on Engineering at Meta.

## Your Algorithm: Giving People More Ways to Personalize Facebook Recommendations

DevFeed: [Your Algorithm: Giving People More Ways to Personalize Facebook Recommendations](<https://devfeed.tech/articles/your-algorithm-giving-people-more-ways-to-personalize-facebook-recommendations-126663.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/11/10/ml-applications/your-algorithm-giving-people-more-ways-to-personalize-facebook-recommendations/>)

Author: Jiarui Wang; Yuxin Cui; Bosen Ding; Zhaohui Guo; James Mason; Facundo Severi; Qingyuan Song; Ziyun Xu

Published: 2026-11-10T17:00:49Z

Content type: article

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>), [Meta ML Applications](<https://devfeed.tech/sources/meta-ml-applications.md>)

Topics: [recommendation systems](<https://devfeed.tech/topics/recommendation-systems.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [AI for interactive applications](<https://devfeed.tech/topics/ai-for-interactive-applications.md>), [User interface design](<https://devfeed.tech/topics/ui-design.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-powered](<https://devfeed.tech/tags/ai-powered.md>), [algorithm](<https://devfeed.tech/tags/algorithm.md>), [facebook](<https://devfeed.tech/tags/facebook.md>), [llms](<https://devfeed.tech/tags/llms.md>), [ml-applications](<https://devfeed.tech/tags/ml-applications.md>), [natural-language](<https://devfeed.tech/tags/natural-language.md>), [recommendations](<https://devfeed.tech/tags/recommendations.md>), [reels](<https://devfeed.tech/tags/reels.md>), [responsive](<https://devfeed.tech/tags/responsive.md>), [system](<https://devfeed.tech/tags/system.md>), [your-algorithm](<https://devfeed.tech/tags/your-algorithm.md>)

### AI overview

Your algorithm lets people state and edit interests for Facebook Feed and Reels, with changes stored in one preference store used by both. Each "Interested In" topic triggers dedicated retrieval and ranking signals, with spacing and daily caps to surface new interests without letting them dominate. "Not Interested In" preferences are enforced through embedding-based filtering and demotion, an LLM reranker, and removal of matching items already cached on the device. Supplementary fetches speed up new-interest recommendations; cache invalidation applies exclusions to queued content. LLMs also generate initial interest profiles, validate edits, and review final recommendations. The system uses separate safety and integrity checks, and Meta identifies ambiguous input and more precise responses as areas for improvement.

### Source excerpt

Your algorithm is a suite of AI-powered tools that makes Facebook recommendations more transparent and steerable, with a single editable view of the interests that shape Feed and Reels. We're sharing how Your algorithm turns people's natural language preferences into responsive recommendations, and how LLMs are tailored to specific tasks within the system. Traditionally, the [...] Read More... The post Your Algorithm: Giving People More Ways to Personalize Facebook Recommendations appeared first on Engineering at Meta.

## How We Built TLX: A Hardware-Native GPU Compiler for Production ML at Scale

DevFeed: [How We Built TLX: A Hardware-Native GPU Compiler for Production ML at Scale](<https://devfeed.tech/articles/how-we-built-tlx-a-hardware-native-gpu-compiler-for-production-ml-at-scale-126664.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/11/10/open-source/tlx-a-hardware-native-gpu-compiler-for-production-ml/>)

Author: Hongtao Yu; Peng Chen; Daohang Shi; Nick Riasanovsky; Lei Wang; Partha Kanuparthy

Published: 2026-11-10T16:59:31Z

Content type: article

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>), [Meta ML Applications](<https://devfeed.tech/sources/meta-ml-applications.md>)

Topics: [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [mlir](<https://devfeed.tech/topics/mlir.md>), [Production Engineering](<https://devfeed.tech/topics/production-engineering.md>), [scheduling](<https://devfeed.tech/topics/scheduling.md>)

Tags: [architectures](<https://devfeed.tech/tags/architectures.md>), [capabilities](<https://devfeed.tech/tags/capabilities.md>), [cluster](<https://devfeed.tech/tags/cluster.md>), [compiler](<https://devfeed.tech/tags/compiler.md>), [developers](<https://devfeed.tech/tags/developers.md>), [devinfra](<https://devfeed.tech/tags/devinfra.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [engineers](<https://devfeed.tech/tags/engineers.md>), [extension](<https://devfeed.tech/tags/extension.md>), [extensions](<https://devfeed.tech/tags/extensions.md>), [first](<https://devfeed.tech/tags/first.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [helps](<https://devfeed.tech/tags/helps.md>), [improvements](<https://devfeed.tech/tags/improvements.md>), [inference](<https://devfeed.tech/tags/inference.md>), [large](<https://devfeed.tech/tags/large.md>), [llm](<https://devfeed.tech/tags/llm.md>), [ml](<https://devfeed.tech/tags/ml.md>), [ml-applications](<https://devfeed.tech/tags/ml-applications.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [production](<https://devfeed.tech/tags/production.md>), [production-engineering](<https://devfeed.tech/tags/production-engineering.md>), [scale](<https://devfeed.tech/tags/scale.md>)

### AI overview

TLX extends Triton with an execution layer that makes warp-group cooperation, synchronization, asynchronous memory movement, and cluster execution explicit while retaining Triton's tensor operations, Python workflow, and autotuning. The authors describe this as a middle ground between CUDA's thread-level control and Triton's more implicit blocked model. TLX provides access to hardware features such as TMA, async tensor-core operations, and TMEM; layouts are inferred by default and can be tuned when needed, which the authors say eases fusion. Reported deployments include GEM training on Blackwell, where a TLX GDPA kernel delivered up to 2x forward speedup; RecSys inference, where a warp-specialized compression kernel achieved 4x speedup on H100 and a co-designed FlashAttention optimization reduced compute-intensive net latency by up to two-thirds; and distributed grouped GEMM for MoE pretraining. These are workload-specific results. The article says TLX targets NVIDIA Hopper and Blackwell and AMD MI300 and MI350X, with broader architecture support and PyTorch 2.x integration still in development.

### Source excerpt

Triton Low-Level Extensions (TLX), is an extension to Triton that helps engineers optimize GPU performance while preserving Triton's productive programming model. TLX introduces explicit warp-group and cluster-level orchestration, letting developers adopt new hardware capabilities without sacrificing velocity. TLX has demonstrated performance improvements in RecSys training, RecSys inference, and large-scale LLM pretraining. Modern GPU architectures increasingly [...] Read More... The post How We Built TLX: A Hardware-Native GPU Compiler for Production ML at Scale appeared first on Engineering at Meta.

## Extending LLDB to Debug Applications That Span CPUs and GPUs

DevFeed: [Extending LLDB to Debug Applications That Span CPUs and GPUs](<https://devfeed.tech/articles/extending-lldb-to-debug-applications-that-span-cpus-and-gpus-126662.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/11/09/developer-tools/extending-lldb-to-debug-applications-that-span-cpus-and-gpus/>)

Author: Greg Clayton; Satyanarayana Janga; Jeffrey Tan; Janet Yang; Bar Soloveychik; Chen Li; Vipul Bansal

Published: 2026-11-09T17:00:12Z

Content type: article

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>)

Topics: [Tool](<https://devfeed.tech/topics/tool.md>), [Developer Tools](<https://devfeed.tech/topics/developer-tools.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-applications](<https://devfeed.tech/tags/ai-applications.md>), [applications](<https://devfeed.tech/tags/applications.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [can](<https://devfeed.tech/tags/can.md>), [control](<https://devfeed.tech/tags/control.md>), [cpu](<https://devfeed.tech/tags/cpu.md>), [cpus](<https://devfeed.tech/tags/cpus.md>), [debug](<https://devfeed.tech/tags/debug.md>), [debugger](<https://devfeed.tech/tags/debugger.md>), [debugging](<https://devfeed.tech/tags/debugging.md>), [design](<https://devfeed.tech/tags/design.md>), [device](<https://devfeed.tech/tags/device.md>), [devinfra](<https://devfeed.tech/tags/devinfra.md>), [first](<https://devfeed.tech/tags/first.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [gpus](<https://devfeed.tech/tags/gpus.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [lldb](<https://devfeed.tech/tags/lldb.md>), [meta](<https://devfeed.tech/tags/meta.md>)

### AI overview

LLDB's accelerator plugin architecture coordinates CPU and GPU debugging in one session while keeping them as separate targets, each with its own architecture, threads, registers, address space, and execution state. A vendor-specific backend in lldb-server translates device APIs into LLDB concepts; shared debugger features can then serve different accelerators while vendors implement hardware-specific behavior. During a session, plugins can set host breakpoints for runtime events, initialize device debugging, provide connection details, and coordinate host and device execution. Support activates only for applications using a supported GPU runtime; CPU-only debugging keeps its existing path without runtime computational overhead. The framework also supports postmortem GPU debugging and includes a mock plugin for testing without specialized hardware. The article says most common infrastructure had been upstreamed to LLVM at the time of writing, with remaining work and broader vendor integrations planned.

### Source excerpt

We built a vendor-neutral accelerator plugin architecture in lldb-server for debugging applications that span CPUs and GPUs. It separates shared debugger infrastructure from hardware-specific device control. The design represents the CPU and GPU as separate debugger processes while presenting them in one debugging session. Modern AI applications rarely execute in one place. A workload can [...] Read More... The post Extending LLDB to Debug Applications That Span CPUs and GPUs appeared first on Engineering at Meta.

## F3 Lightning: How Meta Accelerates Ads ML Features to Production at Scale

DevFeed: [F3 Lightning: How Meta Accelerates Ads ML Features to Production at Scale](<https://devfeed.tech/articles/f3-lightning-how-meta-accelerates-ads-ml-features-to-production-at-scale-114916.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/11/03/ml-applications/f3-lightning-how-meta-accelerates-ads-ml-features-to-production/>)

Author: Xiaohe Dong; Jonathan He; Jesus Francisco Anaya Gonzalez; Purbesh Sahoo

Published: 2026-11-03T17:00:35Z

Content type: article

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>), [Meta ML Applications](<https://devfeed.tech/sources/meta-ml-applications.md>)

Topics: [Feature Engineering](<https://devfeed.tech/topics/feature-engineering.md>), [Production Engineering](<https://devfeed.tech/topics/production-engineering.md>), [Continuous Delivery (CD)](<https://devfeed.tech/topics/continuous-delivery.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>)

Tags: [ads](<https://devfeed.tech/tags/ads.md>), [ai](<https://devfeed.tech/tags/ai.md>), [authoring](<https://devfeed.tech/tags/authoring.md>), [costs](<https://devfeed.tech/tags/costs.md>), [execution](<https://devfeed.tech/tags/execution.md>), [feature](<https://devfeed.tech/tags/feature.md>), [features](<https://devfeed.tech/tags/features.md>), [framework](<https://devfeed.tech/tags/framework.md>), [graph](<https://devfeed.tech/tags/graph.md>), [learning](<https://devfeed.tech/tags/learning.md>), [lightning](<https://devfeed.tech/tags/lightning.md>), [machine](<https://devfeed.tech/tags/machine.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [meta](<https://devfeed.tech/tags/meta.md>), [ml](<https://devfeed.tech/tags/ml.md>), [ml-applications](<https://devfeed.tech/tags/ml-applications.md>), [models](<https://devfeed.tech/tags/models.md>), [production](<https://devfeed.tech/tags/production.md>), [production-engineering](<https://devfeed.tech/tags/production-engineering.md>), [request](<https://devfeed.tech/tags/request.md>), [scale](<https://devfeed.tech/tags/scale.md>), [serving](<https://devfeed.tech/tags/serving.md>)

### AI overview

F3 Lightning lets Meta release, validate, and roll back each ads ML feature independently, avoiding the all-or-nothing releases caused by F3's shared graph. Independent configurations initially increased CPU use because they repeated work. At service startup, Lightning restores shared execution: stable computation identifiers expose duplicate work; a shared, flattened execution plan runs it once; one request buffer improves memory locality and avoids duplicate result storage; and dense indices make runtime access direct. A failing computation marks dependent work as not ready while independent branches continue. The article reports that comparable extraction features reach production about 3.6 times faster than with traditional F3, with no increase in infrastructure compute cost. Features used by live models receive extra testing before release, while unused new features can follow standard validation.

### Source excerpt

Facebook Feature Framework (F3) is Meta's framework for authoring, compiling, and serving the machine learning features underlying ads ranking models, where a single request can require more than 20,000 features. We first presented F3 at AI @ Scale 2020. F3 reduced execution computational costs by compiling all features into a single merged graph, but that [...] Read More... The post F3 Lightning: How Meta Accelerates Ads ML Features to Production at Scale appeared first on Engineering at Meta.

## Catching Bugs Before They Land: JiTTesting's Human-on-the-Loop Approach at Full Scale

DevFeed: [Catching Bugs Before They Land: JiTTesting's Human-on-the-Loop Approach at Full Scale](<https://devfeed.tech/articles/catching-bugs-before-they-land-jittesting-s-human-on-the-loop-approach-at-full-scale-106789.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/10/29/security/catching-bugs-before-they-land-jittestings-human-on-the-loop-approach-at-full-scale/>)

Author: Sreeja Thummala; Sophie Zeng; Mark Harman; Weini Liu; Eric Liu

Published: 2026-10-29T16:00:12Z

Content type: article

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>), [Meta AI Research](<https://devfeed.tech/sources/meta-ai-research.md>), [Meta ML Applications](<https://devfeed.tech/sources/meta-ml-applications.md>)

Topics: [Test automation](<https://devfeed.tech/topics/test-automation.md>), [ai code review](<https://devfeed.tech/topics/ai-code-review.md>), [human review](<https://devfeed.tech/topics/human-review.md>), [Code review](<https://devfeed.tech/topics/code-review.md>)

Tags: [agentic](<https://devfeed.tech/tags/agentic.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [agentic-software-development](<https://devfeed.tech/tags/agentic-software-development.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-research](<https://devfeed.tech/tags/ai-research.md>), [approach](<https://devfeed.tech/tags/approach.md>), [bugs](<https://devfeed.tech/tags/bugs.md>), [generation](<https://devfeed.tech/tags/generation.md>), [human-on-the-loop](<https://devfeed.tech/tags/human-on-the-loop.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [large-language-models](<https://devfeed.tech/tags/large-language-models.md>), [meta](<https://devfeed.tech/tags/meta.md>), [ml-applications](<https://devfeed.tech/tags/ml-applications.md>), [production](<https://devfeed.tech/tags/production.md>), [scale](<https://devfeed.tech/tags/scale.md>), [security-privacy](<https://devfeed.tech/tags/security-privacy.md>), [tests](<https://devfeed.tech/tags/tests.md>)

### AI overview

Meta's Catching JiTTests generates executable tests to expose behavioral differences introduced by code changes, aiming to catch bugs that code-review-based LLM approaches may miss. It runs overnight on spare capacity against diffs ranked as risky. Two workflows generate candidate tests: one treats each diff like a mutant, while the other infers the change's intent and creates tests for plausible implementation risks. Rule-based RubFake and an ensemble of LLM judges rank likely false positives; engineers set review criteria and assess outcomes in aggregate rather than reviewing every test. Meta reports that 589 engineers engaged with the system, with a 60% true-positive rate among reported findings. In the first half of 2026, it helped catch 74 serious privacy bugs before production, and Meta reports a 20% reduction in privacy SEVs during that period.

### Source excerpt

Over the last year we've been sharing our work leveraging large language models to catch bugs on the fly -- a novel approach we call Catching Just-in-Time Tests (JiTTests). Now, we're sharing the results and positive impact we've seen with Catching JiTTests deployed into production. Our paper, "Just-in-Time Catching Test Generation at Meta," fully details [...] Read More... The post Catching Bugs Before They Land: JiTTesting's Human-on-the-Loop Approach at Full Scale appeared first on Engineering at Meta.

## Meta Andromeda: Powering the Next Generation of Ads Retrieval Through Model-System Co-Design

DevFeed: [Meta Andromeda: Powering the Next Generation of Ads Retrieval Through Model-System Co-Design](<https://devfeed.tech/articles/meta-andromeda-powering-the-next-generation-of-ads-retrieval-through-model-system-co-design-105672.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/10/28/ml-applications/meta-andromeda-next-gen-ads-retrieval-gpu-co-design/>)

Author: Haiyu Lu; Shuo Gu; Yang Yang (Tony); Jingwen Ren; Tianchen Ji; Qing Zhang; Wentian Guo; Haocheng Wu; Siqi Zeng; Yunchen Pu

Published: 2026-10-28T16:00:01Z

Content type: article

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>), [Meta ML Applications](<https://devfeed.tech/sources/meta-ml-applications.md>)

Topics: [GPU optimization](<https://devfeed.tech/topics/gpu-optimization.md>), [information retrieval](<https://devfeed.tech/topics/information-retrieval.md>), [model architecture](<https://devfeed.tech/topics/model-architecture.md>), [performance-engineering](<https://devfeed.tech/topics/performance-engineering.md>), [AI Architecture](<https://devfeed.tech/topics/ai-architecture.md>), [system-design-project](<https://devfeed.tech/topics/system-design-project.md>), [Enhancing AI model performance](<https://devfeed.tech/topics/enhancing-ai-model-performance.md>), [quantization](<https://devfeed.tech/topics/quantization.md>)

Tags: [ads](<https://devfeed.tech/tags/ads.md>), [andromeda](<https://devfeed.tech/tags/andromeda.md>), [co-design](<https://devfeed.tech/tags/co-design.md>), [meta](<https://devfeed.tech/tags/meta.md>), [ml-applications](<https://devfeed.tech/tags/ml-applications.md>), [model](<https://devfeed.tech/tags/model.md>), [production-engineering](<https://devfeed.tech/tags/production-engineering.md>), [retrieval](<https://devfeed.tech/tags/retrieval.md>), [system](<https://devfeed.tech/tags/system.md>)

### AI overview

Meta says its Andromeda ads retrieval system combines a mixture-of-experts model with a fused, GPU-based pipeline to score large ad pools efficiently. The model approaches ranking-level performance at roughly 300 times fewer inference FLOPs than the 2024 retrieval model. Fusing model and system operations keeps intermediate data out of high-bandwidth memory; placing targeting-index matching, scoring and top-K selection on one GPU server cuts network traffic by about 100 times. Meta reports production throughput above one billion ads per second per GPU server and a 2.4% increase in Facebook conversions since 2025. In joint Meta-NVIDIA optimization, individual components improved 2-5 times over baseline; INT8 quantization added up to 2 times scoring throughput with accuracy reported on par with 16-bit precision. The post says the full pipeline benchmarks at multiple times current production throughput, while further optimizations and work on Grace Blackwell remain underway.

### Source excerpt

This post describes the next generation of Meta Andromeda, Meta's personalized ads retrieval engine, and how co-designing the model, system, and GPU hardware unlocks a new efficiency frontier for ads retrieval. In production, the system scores every eligible ad with near ranking-level model performance at retrieval scale and surfaces the top candidates at over one [...] Read More... The post Meta Andromeda: Powering the Next Generation of Ads Retrieval Through Model-System Co-Design appeared first on Engineering at Meta.

## PrivacyLib: Privacy Controls at the Data Access Boundary

DevFeed: [PrivacyLib: Privacy Controls at the Data Access Boundary](<https://devfeed.tech/articles/privacylib-privacy-controls-at-the-data-access-boundary-102293.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/10/27/security/privacylib/>)

Author: Sai Nagabhairava; Hannes Roth; Na Chen; Peter Prelich; Jonathan Bergeron

Published: 2026-10-27T16:00:12Z

Content type: article

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>), [Meta AI Research](<https://devfeed.tech/sources/meta-ai-research.md>), [Meta ML Applications](<https://devfeed.tech/sources/meta-ml-applications.md>)

Topics: [Authorization](<https://devfeed.tech/topics/authorization.md>), [Security](<https://devfeed.tech/topics/security.md>), [ai security](<https://devfeed.tech/topics/ai-security.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>)

Tags: [ai-research](<https://devfeed.tech/tags/ai-research.md>), [boundary](<https://devfeed.tech/tags/boundary.md>), [data](<https://devfeed.tech/tags/data.md>), [deep](<https://devfeed.tech/tags/deep.md>), [deep-dive](<https://devfeed.tech/tags/deep-dive.md>), [dive](<https://devfeed.tech/tags/dive.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [first](<https://devfeed.tech/tags/first.md>), [implementing](<https://devfeed.tech/tags/implementing.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [inside](<https://devfeed.tech/tags/inside.md>), [latest](<https://devfeed.tech/tags/latest.md>), [library](<https://devfeed.tech/tags/library.md>), [logic](<https://devfeed.tech/tags/logic.md>), [look](<https://devfeed.tech/tags/look.md>), [meta](<https://devfeed.tech/tags/meta.md>), [ml-applications](<https://devfeed.tech/tags/ml-applications.md>), [post](<https://devfeed.tech/tags/post.md>), [privacy](<https://devfeed.tech/tags/privacy.md>), [processing](<https://devfeed.tech/tags/processing.md>), [security-privacy](<https://devfeed.tech/tags/security-privacy.md>), [system](<https://devfeed.tech/tags/system.md>), [systems](<https://devfeed.tech/tags/systems.md>)

### AI overview

PrivacyLib moves privacy checks to a shared library inside the clients of hundreds of Meta data and processing systems. Each system integrates once; reusable observing or enforcing modules run at common read, write, RPC, request-entry, and inline-execution points. Checks use local request context and cached configuration to operate at billions per second with low latency. For supported direct RPC paths, ServiceRouter validates a signed token from a successful client check or applies a stricter, less-informed fallback; paths outside those boundaries need their own integration. Annotation changes are simulated against observed flows before release and then reach services through normal deployments, while modules can be enabled gradually with runtime controls. For internal agents on supported paths, AgentLib propagates session context to PrivacyLib, where modules can record accesses or block disallowed operations using destination annotations and agent context.

### Source excerpt

This latest deep dive into Meta's Privacy Aware Infrastructure (PAI) takes a look at PrivacyLib. A privacy requirement must hold as data moves through Meta's diverse data and processing systems. Historically, supporting a new requirement meant implementing the same enforcement logic separately in each system. PrivacyLib replaces this with a common library that runs inside [...] Read More... The post PrivacyLib: Privacy Controls at the Data Access Boundary appeared first on Engineering at Meta.

## Python Typing Survey 2026: AI Tools and the Shifting Checker Landscape

DevFeed: [Python Typing Survey 2026: AI Tools and the Shifting Checker Landscape](<https://devfeed.tech/articles/python-typing-survey-2026-ai-tools-and-the-shifting-checker-landscape-104402.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/10/27/developer-tools/python-typing-survey-2026-ai-tools-checker-landscape/>)

Author: Chris Wiltz

Published: 2026-10-27T15:59:24Z

Content type: news

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>)

Topics: [Python](<https://devfeed.tech/topics/python.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [Tool](<https://devfeed.tech/topics/tool.md>), [Developer experience](<https://devfeed.tech/topics/developer-experience.md>), [Pydantic](<https://devfeed.tech/topics/pydantic.md>)

Tags: [2](<https://devfeed.tech/tags/2.md>), [2026](<https://devfeed.tech/tags/2026.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-tools](<https://devfeed.tech/tags/ai-tools.md>), [but](<https://devfeed.tech/tags/but.md>), [culture](<https://devfeed.tech/tags/culture.md>), [developers](<https://devfeed.tech/tags/developers.md>), [devinfra](<https://devfeed.tech/tags/devinfra.md>), [first](<https://devfeed.tech/tags/first.md>), [half](<https://devfeed.tech/tags/half.md>), [hint](<https://devfeed.tech/tags/hint.md>), [information](<https://devfeed.tech/tags/information.md>), [library](<https://devfeed.tech/tags/library.md>), [little](<https://devfeed.tech/tags/little.md>), [meta](<https://devfeed.tech/tags/meta.md>), [one](<https://devfeed.tech/tags/one.md>), [python](<https://devfeed.tech/tags/python.md>), [responses](<https://devfeed.tech/tags/responses.md>), [selection](<https://devfeed.tech/tags/selection.md>), [survey](<https://devfeed.tech/tags/survey.md>), [tools](<https://devfeed.tech/tags/tools.md>), [type](<https://devfeed.tech/tags/type.md>)

### AI overview

Across 1,794 survey responses, frequent Python type hint use stayed near 86% over three years, while checker choice shifted and developers increasingly mixed tools. Missing or inadequate third party library types remained the leading difficulty (65.1%); 49.9% selected at least two checkers, and 42.7% of multi-checker users reported inconsistencies. Speed was the leading reason for checker choice. AI use in typing grew: 42.5% said AI tools wrote type hints, and some respondents described tools running checkers and fixing errors. But 29.8% were unsure whether types improve AI-assisted coding, and the article says evidence about helpful interactions is still needed. The authors point to better library annotations, clearer editor feedback and documentation, low-configuration first checker runs, and simpler callable syntax as practical areas for improvement. Survey findings reflect respondents, most of whom had at least five years of Python experience.

### Source excerpt

The 2026 Python Typing Survey received 1,794 responses. Type hint use changed little across three years, but checker selection changed sharply. Among respondents who reported which checkers they use, half selected more than one, and missing library type information was the most common difficulty. For the third consecutive year, we asked Python developers how they [...] Read More... The post Python Typing Survey 2026: AI Tools and the Shifting Checker Landscape appeared first on Engineering at Meta.

## Meta Networking at the 2026 OCP Global Summit: Open Innovation for Scale-Up, Scale-Out, and Scale-Across AI Networks

DevFeed: [Meta Networking at the 2026 OCP Global Summit: Open Innovation for Scale-Up, Scale-Out, and Scale-Across AI Networks](<https://devfeed.tech/articles/meta-networking-at-the-2026-ocp-global-summit-open-innovation-for-scale-up-scale-out-and-scale-across-ai-networks-100675.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/10/26/open-source/ocp-summit-2026-meta-networking-talks/>)

Author: Ravindra Sunkad; Omar Baldonado

Published: 2026-10-26T16:00:07Z

Content type: news

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>), [Meta ML Applications](<https://devfeed.tech/sources/meta-ml-applications.md>)

Topics: [AI Networking](<https://devfeed.tech/topics/ai-networking.md>), [rdma](<https://devfeed.tech/topics/rdma.md>), [Open Source](<https://devfeed.tech/topics/open-source.md>), [NVIDIA Spectrum-X Ethernet](<https://devfeed.tech/topics/nvidia-spectrum-x-ethernet.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>), [Test-driven development](<https://devfeed.tech/topics/tdd.md>)

Tags: [102-4t](<https://devfeed.tech/tags/102-4t.md>), [2026](<https://devfeed.tech/tags/2026.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-models](<https://devfeed.tech/tags/ai-models.md>), [ai-networks](<https://devfeed.tech/tags/ai-networks.md>), [broadcom](<https://devfeed.tech/tags/broadcom.md>), [cluster](<https://devfeed.tech/tags/cluster.md>), [data-center-engineering](<https://devfeed.tech/tags/data-center-engineering.md>), [data-infrastructure](<https://devfeed.tech/tags/data-infrastructure.md>), [devinfra](<https://devfeed.tech/tags/devinfra.md>), [global](<https://devfeed.tech/tags/global.md>), [innovation](<https://devfeed.tech/tags/innovation.md>), [meta](<https://devfeed.tech/tags/meta.md>), [ml-applications](<https://devfeed.tech/tags/ml-applications.md>), [networking](<https://devfeed.tech/tags/networking.md>), [networking-traffic](<https://devfeed.tech/tags/networking-traffic.md>), [networks](<https://devfeed.tech/tags/networks.md>), [ocp](<https://devfeed.tech/tags/ocp.md>), [open](<https://devfeed.tech/tags/open.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [scale](<https://devfeed.tech/tags/scale.md>), [scale-across](<https://devfeed.tech/tags/scale-across.md>), [scale-out](<https://devfeed.tech/tags/scale-out.md>), [scale-up](<https://devfeed.tech/tags/scale-up.md>), [summit](<https://devfeed.tech/tags/summit.md>), [up](<https://devfeed.tech/tags/up.md>)

### AI overview

Meta's 2026 OCP Summit overview presents an open, full-stack network strategy for AI clusters spanning scale-up, scale-out, and scale-across. Highlights include IcePack/IceCube liquid-cooled racks with 102.4T switches, and a flattened scale-out fabric that connects each GPU to eight planes and reduces a three-tier Clos to two tiers using NIC packet spraying and joint host-switch control. For cross-site capacity, BAG links buildings, campuses, and regions, alongside SRv6 backbones running on FBOSS and SAI. Scale-up work includes Tahan, an Ethernet switch for MTIA 300, and ESUN, whose 1.0 specification is released and 1.1 is in progress. Meta also describes OSPA, a multi-vendor switch-platform standard submitted to OCP, and open-sourcing MetaRoCE software, beginning with a user-space implementation; a DPDK version is planned. The excerpt lists talks on optics, interconnect reliability, deployment, and operation, but does not provide detailed results for most of them.

### Source excerpt

At the 2026 Open Compute Project (OCP) Global Summit, Meta shared the networking innovations that are directly supporting our latest AI models, including Muse Wonder and Muse Spark, and our gigawatt-scale Prometheus cluster. Highlights IcePack & IceCube: Meta's first liquid-cooled network rack with 102.4T switch blades (128x800G, Broadcom Tomahawk 6C and NVIDIA Spectrum-6 variants), running [...] Read More... The post Meta Networking at the 2026 OCP Global Summit: Open Innovation for Scale-Up, Scale-Out, and Scale-Across AI Networks appeared first on Engineering at Meta.

## GenRecs Tokenization: Transforming Ads Recommendation at Meta with Stable Model Vocabulary

DevFeed: [GenRecs Tokenization: Transforming Ads Recommendation at Meta with Stable Model Vocabulary](<https://devfeed.tech/articles/genrecs-tokenization-transforming-ads-recommendation-at-meta-with-stable-model-vocabulary-90425.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/10/19/ml-applications/genrecs-tokenization-transforming-ads-recommendation-at-meta-with-stable-model-vocabulary/>)

Author: Carl Hu; Justin Johnson; Parish Aggarwal; Kaushik Rangadurai; Senthil Manickavelu; Tianwei Zhang

Published: 2026-10-19T16:00:47Z

Content type: article

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>), [Meta ML Applications](<https://devfeed.tech/sources/meta-ml-applications.md>)

Topics: [tokenization](<https://devfeed.tech/topics/tokenization.md>), [personalization](<https://devfeed.tech/topics/personalization.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [quantization](<https://devfeed.tech/topics/quantization.md>)

Tags: [data](<https://devfeed.tech/tags/data.md>), [llms](<https://devfeed.tech/tags/llms.md>), [meta](<https://devfeed.tech/tags/meta.md>), [ml-applications](<https://devfeed.tech/tags/ml-applications.md>), [model](<https://devfeed.tech/tags/model.md>), [reasoning](<https://devfeed.tech/tags/reasoning.md>), [recommendation](<https://devfeed.tech/tags/recommendation.md>), [user](<https://devfeed.tech/tags/user.md>)

### AI overview

Meta describes GenRecs Tokenization, a semantic representation for ads and other items that replaces individual item IDs with sequences of tokens from a stable vocabulary. It combines item content with engagement signals so recommendation models can generalize to new or low-engagement items. Meta reports a 0.35% conversion lift when the tokenizer was paired with its Generative Ads Model and Adaptive Ranking Model, based on internal data measured in February 2026.

### Source excerpt

Generative Recommenders (GenRecs) represent a paradigm shift from scoring every ad to reasoning about ad content and user preferences together in order to predict the best outcome for advertisers. Leveraging the world knowledge built into LLMs makes ad matching more intelligent and precise, compounding gains for advertisers. GenRecs Tokenization is the foundational data representation that [...] Read More... The post GenRecs Tokenization: Transforming Ads Recommendation at Meta with Stable Model Vocabulary appeared first on Engineering at Meta.

## The Thin Orchestrator: How We Made Our AI Agents More Reliable by Giving the Router Less to Do

DevFeed: [The Thin Orchestrator: How We Made Our AI Agents More Reliable by Giving the Router Less to Do](<https://devfeed.tech/articles/the-thin-orchestrator-how-we-made-our-ai-agents-more-reliable-by-giving-the-router-less-to-do-89999.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/10/19/ml-applications/thin-orchestrator-ai-agent-architecture/>)

Author: Plawan Rath; Akhil Jindal; Ashwini Verma

Published: 2026-10-19T15:30:37Z

Content type: article

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>), [Meta AI Research](<https://devfeed.tech/sources/meta-ai-research.md>), [Meta ML Applications](<https://devfeed.tech/sources/meta-ml-applications.md>)

Topics: [agent orchestration](<https://devfeed.tech/topics/agent-orchestration.md>), [Multi Agent Systems](<https://devfeed.tech/topics/multi-agent-systems.md>), [supervisor agent pattern](<https://devfeed.tech/topics/supervisor-agent-pattern.md>), [agent to agent communication](<https://devfeed.tech/topics/agent-to-agent-communication.md>), [Test automation](<https://devfeed.tech/topics/test-automation.md>)

Tags: [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [ai-research](<https://devfeed.tech/tags/ai-research.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [ml-applications](<https://devfeed.tech/tags/ml-applications.md>), [orchestration](<https://devfeed.tech/tags/orchestration.md>), [reliability](<https://devfeed.tech/tags/reliability.md>)

### AI overview

Meta describes a production AI-agent architecture for onboarding data columns across recommendation-system pipelines. A deterministic, thin orchestrator routes requests and applies reliability checks, while 14 specialized agents handle domain reasoning in isolated contexts. The article reports that configuration preparation fell from about a day with an earlier system to under an hour, and that roughly two-thirds of changes land without significant human edits. It also explains how reusable workflow templates and runtime-loaded knowledge support new use cases.

### Source excerpt

Sometimes, giving an AI agent all of the knowledge it needs is the wrong approach. At Meta, we've seen notable improvements in efficiency and accuracy by keeping the orchestration layer of our data pipeline agents deliberately thin: It routes work and enforces reliability checks, while specialized agents do the domain reasoning. We're sharing our own [...] Read More... The post The Thin Orchestrator: How We Made Our AI Agents More Reliable by Giving the Router Less to Do appeared first on Engineering at Meta.

## Inside Prometheus: Meta's Journey to a Gigawatt-Scale Supercluster

DevFeed: [Inside Prometheus: Meta's Journey to a Gigawatt-Scale Supercluster](<https://devfeed.tech/articles/inside-prometheus-meta-s-journey-to-a-gigawatt-scale-supercluster-89979.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/10/19/data-center-engineering/inside-prometheus-meta-gigawatt-scale-supercluster/>)

Author: Kevin Lee; Yashar Bayani; Ahmed Sharif; Yee Jiun Song

Published: 2026-10-19T13:00:10Z

Content type: article

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>), [Meta ML Applications](<https://devfeed.tech/sources/meta-ml-applications.md>)

Topics: [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [NVLink](<https://devfeed.tech/topics/nvlink.md>), [GB200](<https://devfeed.tech/topics/gb200.md>), [Vera Rubin NVL72](<https://devfeed.tech/topics/vera-rubin-nvl72.md>), [clean energy data centers](<https://devfeed.tech/topics/clean-energy-data-centers.md>), [Cloud Architecture](<https://devfeed.tech/topics/cloud-architecture.md>)

Tags: [400g](<https://devfeed.tech/tags/400g.md>), [51-2t](<https://devfeed.tech/tags/51-2t.md>), [core-infra](<https://devfeed.tech/tags/core-infra.md>), [data-center-engineering](<https://devfeed.tech/tags/data-center-engineering.md>), [data-infrastructure](<https://devfeed.tech/tags/data-infrastructure.md>), [gpus](<https://devfeed.tech/tags/gpus.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [meta](<https://devfeed.tech/tags/meta.md>), [ml-applications](<https://devfeed.tech/tags/ml-applications.md>), [networking-traffic](<https://devfeed.tech/tags/networking-traffic.md>), [production-engineering](<https://devfeed.tech/tags/production-engineering.md>), [prometheus](<https://devfeed.tech/tags/prometheus.md>), [release](<https://devfeed.tech/tags/release.md>), [scale](<https://devfeed.tech/tags/scale.md>), [scaling](<https://devfeed.tech/tags/scaling.md>)

### AI overview

Meta describes Prometheus, a gigawatt-scale AI supercluster in New Albany, Ohio, and its plan to expand toward roughly one million GPUs. The article details its GPU platforms, networking, reliability and fault-recovery systems, power management, and liquid-cooling approaches. It reports that engineering changes reduced interruptions by 99% from a peak of nearly 1,000 events per 8K GPUs daily.

### Source excerpt

With the recent release of the Muse family of foundational models, we want to take you inside the infrastructure that has made it all possible. The first phase of Prometheus, our gigawatt-scale supercluster, is already complete and we're working toward scaling it to comprise ~1 million GPUs. Prometheus also features technologies like closed-loop liquid cooling [...] Read More... The post Inside Prometheus: Meta's Journey to a Gigawatt-Scale Supercluster appeared first on Engineering at Meta.

## How Meta Built Forum With AI-Native Engineering

DevFeed: [How Meta Built Forum With AI-Native Engineering](<https://devfeed.tech/articles/how-meta-built-forum-with-ai-native-engineering-88235.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/10/15/ml-applications/how-meta-built-forum-ai-native-engineering/>)

Author: Shengbo Guo; Bohou Luo; Shubhojeet Sarkar; Vivian Qu; Max (Xiangjun) Fan; Prerna Jain; Mike Plumpe; Chinmay Abhyankar; Rish Tandon

Published: 2026-10-15T19:00:03Z

Content type: article

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>), [Meta ML Applications](<https://devfeed.tech/sources/meta-ml-applications.md>)

Topics: [Loop Engineering](<https://devfeed.tech/topics/loop-engineering.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>), [engineering-culture](<https://devfeed.tech/topics/engineering-culture.md>), [Chaos Engineering](<https://devfeed.tech/topics/chaos-engineering.md>)

Tags: [agentic](<https://devfeed.tech/tags/agentic.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-native-engineering](<https://devfeed.tech/tags/ai-native-engineering.md>), [android](<https://devfeed.tech/tags/android.md>), [continuous-verification](<https://devfeed.tech/tags/continuous-verification.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [ios](<https://devfeed.tech/tags/ios.md>), [ml-applications](<https://devfeed.tech/tags/ml-applications.md>), [web](<https://devfeed.tech/tags/web.md>)

### AI overview

Meta describes how a small team built Forum, a standalone Facebook Groups app, by integrating AI agents across planning, implementation, testing, verification, and post-launch investigation. Engineers retained responsibility for product decisions and sign-off, while automated checks and simulated user testing supported agent-run development cycles. The article also describes adapting the codebase and build setup, and reports that stress testing found a memory issue before launch and that Forum reached core feature parity with Facebook Groups in eight weeks.

### Source excerpt

Meta's newest app, Forum, is for anyone who wants to go deeper into the conversations in their Facebook groups. To build Forum, Meta engineers adopted a whole new way of working by bringing an AI agent into the development cycle. We're sharing insights into what it took to build Forum with agentic AI: how we [...] Read More... The post How Meta Built Forum With AI-Native Engineering appeared first on Engineering at Meta.

## 2026 World Cup: Breaking Records, Spiky Charts and Reliability at Scale

DevFeed: [2026 World Cup: Breaking Records, Spiky Charts and Reliability at Scale](<https://devfeed.tech/articles/2026-world-cup-breaking-records-spiky-charts-and-reliability-at-scale-88236.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/10/15/production-engineering/2026-world-cup-record-traffic-reliability-at-scale/>)

Author: Syamla Bandla; Vincent Mauge; Michael Stella; Deniz Hastorun; Francois Richard; Javier Frias; Mike McCall

Published: 2026-10-15T18:30:29Z

Content type: article

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>)

Topics: [Production Engineering](<https://devfeed.tech/topics/production-engineering.md>), [API Business Models](<https://devfeed.tech/topics/api-business-models.md>)

Tags: [autoscaling](<https://devfeed.tech/tags/autoscaling.md>), [capacity-planning](<https://devfeed.tech/tags/capacity-planning.md>), [connectivity](<https://devfeed.tech/tags/connectivity.md>), [culture](<https://devfeed.tech/tags/culture.md>), [data-infrastructure](<https://devfeed.tech/tags/data-infrastructure.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [instagram](<https://devfeed.tech/tags/instagram.md>), [networking-traffic](<https://devfeed.tech/tags/networking-traffic.md>), [production-engineering](<https://devfeed.tech/tags/production-engineering.md>), [reliability](<https://devfeed.tech/tags/reliability.md>), [scale](<https://devfeed.tech/tags/scale.md>), [stress-test](<https://devfeed.tech/tags/stress-test.md>), [whatsapp](<https://devfeed.tech/tags/whatsapp.md>)

### AI overview

Meta describes how its infrastructure handled record traffic during the 2026 World Cup, including messaging, video, and recommendation activity. It attributes the traffic peaks chiefly to breaks and post-match moments, and details capacity buffers, targeted infrastructure upgrades, load testing, graceful degradation, and regional demand forecasting. Meta also says teams continued normal deployments throughout the event.

### Source excerpt

Every four years, the World Cup becomes the largest planned stress test of our infrastructure. This year it also set an all-time record! There are a few moments throughout the year, like New Years Eve, that really let us see the collective impact of everything we've built -- but none are bigger than the quadrennial [...] Read More... The post 2026 World Cup: Breaking Records, Spiky Charts and Reliability at Scale appeared first on Engineering at Meta.

## Sapienz Receives ISSTA 2026 Impact Paper Award

DevFeed: [Sapienz Receives ISSTA 2026 Impact Paper Award](<https://devfeed.tech/articles/sapienz-receives-issta-2026-impact-paper-award-88234.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/10/15/developer-tools/sapienz-issta-2026-impact-paper-award-meta/>)

Author: Chris Wiltz

Published: 2026-10-15T16:00:59Z

Content type: news

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>)

Topics: [Test automation](<https://devfeed.tech/topics/test-automation.md>), [Android Testing](<https://devfeed.tech/topics/android-testing.md>), [Software Engineering](<https://devfeed.tech/topics/software-engineering.md>), [mutation-testing](<https://devfeed.tech/topics/mutation-testing.md>)

Tags: [android](<https://devfeed.tech/tags/android.md>), [automated-testing](<https://devfeed.tech/tags/automated-testing.md>), [culture](<https://devfeed.tech/tags/culture.md>), [devinfra](<https://devfeed.tech/tags/devinfra.md>), [impact](<https://devfeed.tech/tags/impact.md>), [meta](<https://devfeed.tech/tags/meta.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [paper](<https://devfeed.tech/tags/paper.md>), [software-testing](<https://devfeed.tech/tags/software-testing.md>)

### AI overview

Sapienz, a tool for automated Android app testing, received the ISSTA 2026 Most Influential Paper Award for research published in 2016. Its multi-objective testing found 558 previously unknown crashes in an evaluation of 1,000 Google Play apps, and the system later became testing infrastructure at Meta. It now runs millions of tests daily across Meta platforms, while newer LLM agents extend app exploration toward functional correctness and automated repair.

### Source excerpt

Ke Mao, Mark Harman, and Yue Jia have been recognized with the ACM SIGSOFT International Symposium on Software Testing and Analysis (ISSTA) 2026 Most Influential Paper Award for their 2016 research, "Sapienz: multi-objective automated testing for Android applications." The recognition celebrates a decade of impact from work that began at University College London and went [...] Read More... The post Sapienz Receives ISSTA 2026 Impact Paper Award appeared first on Engineering at Meta.

## Meta's Generative Ads Model (GEM): Scaling the Central Model and Compute Engine for Frontier Ads AI

DevFeed: [Meta's Generative Ads Model (GEM): Scaling the Central Model and Compute Engine for Frontier Ads AI](<https://devfeed.tech/articles/meta-s-generative-ads-model-gem-scaling-the-central-model-and-compute-engine-for-frontier-ads-ai-86364.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/10/13/ml-applications/generative-ads-model-gem-scaling-frontier-ads-ai/>)

Author: Huayu Li; Darren Liu; Xiaoyi (Leo) Liu; Ellie Wen; Jackie (Jiaqi) Xu; Chunzhi Yang; Reva Srinivasan

Published: 2026-10-13T16:01:00Z

Content type: article

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>), [Meta ML Applications](<https://devfeed.tech/sources/meta-ml-applications.md>)

Topics: [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [AI Integration Strategies](<https://devfeed.tech/topics/ai-integration-strategies.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [ml-applications](<https://devfeed.tech/tags/ml-applications.md>), [recsys](<https://devfeed.tech/tags/recsys.md>)

### AI overview

Meta describes GEM, a unified recommendation foundation model for ads that predicts clicks and conversions across its advertising surfaces. The article outlines its Kunlun architecture, methods for transferring model knowledge to lower-latency serving models, and techniques for training at LLM scale across thousands of GPUs. Meta reports doubled end-to-end training efficiency to 20-25% model FLOPS utilization while scaling total training FLOPs by 4x.

### Source excerpt

Last year, we introduced GEM, the industry's largest recommendation system (RecSys) foundation model. GEM established Meta's first LLM-inspired paradigm for ads recommendation, driving major performance gains. Building on this foundation, our next goal is to further scale both compute and intelligence transfer with next-generation hardware, and expand this capability across Meta's entire ads ecosystem. This [...] Read More... The post Meta's Generative Ads Model (GEM): Scaling the Central Model and Compute Engine for Frontier Ads AI appeared first on Engineering at Meta.

## OCP Summit 2026: Powering AI with Open Innovations in Scale-Up, Scale-Out, and Scale-Across Networks

DevFeed: [OCP Summit 2026: Powering AI with Open Innovations in Scale-Up, Scale-Out, and Scale-Across Networks](<https://devfeed.tech/articles/ocp-summit-2026-powering-ai-with-open-innovations-in-scale-up-scale-out-and-scale-across-networks-85618.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/10/12/open-source/ocp-summit-2026-powering-ai-with-open-innovations-in-scale-up-scale-out-and-scale-across-networks/>)

Author: Chris Wiltz

Published: 2026-10-12T21:19:18Z

Content type: article

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>), [Meta ML Applications](<https://devfeed.tech/sources/meta-ml-applications.md>)

Topics: [AI Networking](<https://devfeed.tech/topics/ai-networking.md>), [Cloud Networking](<https://devfeed.tech/topics/cloud-networking.md>), [Distributed Systems & Parallel Computing](<https://devfeed.tech/topics/distributed-systems-parallel-computing.md>), [Open Source](<https://devfeed.tech/topics/open-source.md>)

Tags: [ai-clusters](<https://devfeed.tech/tags/ai-clusters.md>), [ai-infrastructure](<https://devfeed.tech/tags/ai-infrastructure.md>), [ai-networking](<https://devfeed.tech/tags/ai-networking.md>), [arm](<https://devfeed.tech/tags/arm.md>), [data-center-engineering](<https://devfeed.tech/tags/data-center-engineering.md>), [data-infrastructure](<https://devfeed.tech/tags/data-infrastructure.md>), [devinfra](<https://devfeed.tech/tags/devinfra.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [ml-applications](<https://devfeed.tech/tags/ml-applications.md>), [networking-traffic](<https://devfeed.tech/tags/networking-traffic.md>), [open-source](<https://devfeed.tech/tags/open-source.md>)

### AI overview

Meta outlines networking infrastructure and open standards it is presenting at OCP Summit 2026, including liquid-cooled 102.4T switches, a multiplane fabric for large AI clusters, and a multi-site aggregation network. The article also describes Ethernet scale-up work, the Open Switch Platform Architecture initiative, and open-source FBOSS and MetaRoCE implementations.

### Source excerpt

At the 2026 Open Compute Project Global Summit, Meta is sharing the next generation of open networking infrastructure, including: Our first liquid-cooled network rack (IcePack), and 102.4T switch platform (IceCube). Flattened Planar Fabric (FPF), a multiplane backend scale-out fabric that fundamentally changes how GPUs connect to the network MetaRoCE, our open-source AI transport protocol. Launching [...] Read More... The post OCP Summit 2026: Powering AI with Open Innovations in Scale-Up, Scale-Out, and Scale-Across Networks appeared first on Engineering at Meta.

## Shipping Media Over QUIC at Meta Scale for Instagram Live Ingest

DevFeed: [Shipping Media Over QUIC at Meta Scale for Instagram Live Ingest](<https://devfeed.tech/articles/shipping-media-over-quic-at-meta-scale-for-instagram-live-ingest-85619.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/10/12/video-engineering/shipping-media-over-quic-at-meta-scale-for-instagram-live-ingest/>)

Author: Konstantin Tsoy; Aman Sharma; Hari Anil Kumar; Alan Frindell; Sandarsh Srivastava

Published: 2026-10-12T16:01:32Z

Content type: article

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>)

Topics: [QUIC](<https://devfeed.tech/topics/quic.md>)

Tags: [android](<https://devfeed.tech/tags/android.md>), [data-infrastructure](<https://devfeed.tech/tags/data-infrastructure.md>), [instagram](<https://devfeed.tech/tags/instagram.md>), [ios](<https://devfeed.tech/tags/ios.md>), [meta](<https://devfeed.tech/tags/meta.md>), [networking-traffic](<https://devfeed.tech/tags/networking-traffic.md>), [quic](<https://devfeed.tech/tags/quic.md>), [real-time](<https://devfeed.tech/tags/real-time.md>), [streaming-video](<https://devfeed.tech/tags/streaming-video.md>), [video](<https://devfeed.tech/tags/video.md>), [video-engineering](<https://devfeed.tech/tags/video-engineering.md>), [web](<https://devfeed.tech/tags/web.md>)

### AI overview

Meta describes deploying Media over QUIC (MoQ) for Instagram Live video ingest on iOS and Android. The rollout matched the existing RUSH protocol on streaming quality and reliability measures, and Meta plans to explore broader live video delivery and playback uses.

### Source excerpt

Meta has begun deploying Media over QUIC (MoQ), an emerging open protocol for real-time video streaming at scale. MoQ will help simplify our protocol stack for live streaming video and enhance overall user experience for live video. We're sharing how we shipped MoQ on Instagram Live ingest, and what the future looks like from here. [...] Read More... The post Shipping Media Over QUIC at Meta Scale for Instagram Live Ingest appeared first on Engineering at Meta.

## Turning Opportunistic AI Inference Capacity Into Better Ads Recommendations

DevFeed: [Turning Opportunistic AI Inference Capacity Into Better Ads Recommendations](<https://devfeed.tech/articles/turning-opportunistic-ai-inference-capacity-into-better-ads-recommendations-85617.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/10/12/ml-applications/turning-opportunistic-ai-inference-capacity-into-better-ads-recommendations/>)

Author: Jinfu Leng; Rohith Menon; Nicolas Stier; Praveen Sampath; Hunter Zhang; Xiaocong Liang; Haonan Wu; Michael Fulthorp; Tao Cai; Paras M

Published: 2026-10-12T16:00:50Z

Content type: article

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>), [Meta ML Applications](<https://devfeed.tech/sources/meta-ml-applications.md>)

Topics: [recommendation systems](<https://devfeed.tech/topics/recommendation-systems.md>), [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [Distributed Systems & Parallel Computing](<https://devfeed.tech/topics/distributed-systems-parallel-computing.md>), [High-Performance Computing](<https://devfeed.tech/topics/high-performance-computing.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-inference](<https://devfeed.tech/tags/ai-inference.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [ml-applications](<https://devfeed.tech/tags/ml-applications.md>), [recommendations](<https://devfeed.tech/tags/recommendations.md>), [scale](<https://devfeed.tech/tags/scale.md>)

### AI overview

Meta describes a system that uses locally idle, preemptible inference capacity to improve ads recommendations while protecting baseline serving. It combines Rank More Ads, which expands candidate ranking using CPU and accelerator capacity, with Serve Better Models, which routes selected requests to stronger models using accelerator surplus. Isolation, feedback controllers, and fast value-aware load shedding let the optional workloads adapt as capacity changes. The article also outlines plans to broaden the system's elasticity and automate capacity allocation.

### Source excerpt

At Meta, we've built a system that enhances ads recommendations by leveraging idle inference capacity in our fleet. Two elasticity levers in the system allow ads recommendation work to expand across opportunistic capacity, without compromising baseline performance. This post describes how we designed and now operate the system at Meta's scale and how it drives [...] Read More... The post Turning Opportunistic AI Inference Capacity Into Better Ads Recommendations appeared first on Engineering at Meta.

## NTS: Authenticated Time at Meta

DevFeed: [NTS: Authenticated Time at Meta](<https://devfeed.tech/articles/nts-authenticated-time-at-meta-65891.md>)

Original publisher: [Read original article](<https://engineering.fb.com/2026/10/06/production-engineering/nts-authenticated-time-at-meta/>)

Author: Oleg Obleukhov

Published: 2026-10-06T16:00:06Z

Content type: article

Language: en

Sources: [Engineering at Meta](<https://devfeed.tech/sources/engineering-at-meta.md>)

Topics: [Cryptography](<https://devfeed.tech/topics/cryptography.md>), [enterprise certificate authority certificates](<https://devfeed.tech/topics/enterprise-certificate-authority-certificates.md>), [Rfc 868](<https://devfeed.tech/topics/rfc-868.md>)

Tags: [meta](<https://devfeed.tech/tags/meta.md>), [network](<https://devfeed.tech/tags/network.md>), [networking-traffic](<https://devfeed.tech/tags/networking-traffic.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [production-engineering](<https://devfeed.tech/tags/production-engineering.md>), [rfc](<https://devfeed.tech/tags/rfc.md>), [security](<https://devfeed.tech/tags/security.md>), [time](<https://devfeed.tech/tags/time.md>)

### AI overview

Meta describes its public Network Time Security service, which authenticates time packets and uses stateless cookies derived from a shared secret. The article explains the NTS key exchange and authenticated NTP flow, its security limits, and the lack of built-in NTS support in major mobile time clients.

### Source excerpt

Meta's public time service now speaks NTS (Network Time Security, RFC 8915) at nts.meta.com. Packets are authenticated, so a device can verify the time came from us and was not modified on the way. Our NTS servers hold no per-client state. Cookie keys are derived, not stored and not replicated. We've open sourced everything, including [...] Read More... The post NTS: Authenticated Time at Meta appeared first on Engineering at Meta.

[Next page](<https://devfeed.tech/sources/engineering-at-meta.md?cursor=WyIyMDI2LTEwLTA2VDE2OjAwOjA2KzAwOjAwIiwgIjJlZDg5M2VmLTgxMDAtNDkxMS04YmI3LTAyNmRkNGE0MzMyNSJd>)