Reading room.
Engineering blogs & AI research
A reading room from engineering teams and AI labs: titles, short feed excerpts, and links to the originals. For my own essays, explore the writing archive.
- 1359 articles tracked
- 37 sources
- Refresh scheduled daily
Sources
Your Algorithm: Giving People More Ways to Personalize Facebook Recommendations
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 t…
How We Built TLX: A Hardware-Native GPU Compiler for Production ML at Scale
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 orch…
Extending LLDB to Debug Applications That Span CPUs and GPUs
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 r…
F3 Lightning: How Meta Accelerates Ads ML Features to Production at Scale
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 firs…
Catching Bugs Before They Land: JiTTesting’s Human-on-the-Loop Approach at Full Scale
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 impa…
Meta Andromeda: Powering the Next Generation of Ads Retrieval Through Model-System Co-Design
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 produc…
Python Typing Survey 2026: AI Tools and the Shifting Checker Landscape
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 t…
PrivacyLib: Privacy Controls at the Data Access Boundary
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…
Meta Networking at the 2026 OCP Global Summit: Open Innovation for Scale-Up, Scale-Out, and Scale-Across AI Networks
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 clus…
Inside Prometheus: Meta’s Journey to a Gigawatt-Scale Supercluster
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 c…
GenRecs Tokenization: Transforming Ads Recommendation at Meta with Stable Model Vocabulary
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 kn…
The Thin Orchestrator: How We Made Our AI Agents More Reliable by Giving the Router Less to Do
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 delibe…
Sapienz Receives ISSTA 2026 Impact Paper Award
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-objec…
How Meta Built Forum With AI-Native Engineering
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 developme…
2026 World Cup: Breaking Records, Spiky Charts and Reliability at Scale
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…
Meta’s Generative Ads Model (GEM): Scaling the Central Model and Compute Engine for Frontier Ads AI
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 o…
Turning Opportunistic AI Inference Capacity Into Better Ads Recommendations
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 capacit…
OCP Summit 2026: Powering AI with Open Innovations in Scale-Up, Scale-Out, and Scale-Across Networks
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). Flat…
Shipping Media Over QUIC at Meta Scale for Instagram Live Ingest
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…
NTS: Authenticated Time at Meta
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 n…
Bringing Private Processing to Meta AI Glasses
We believe glasses are the best form factor for having AI help throughout your day. They can understand your personal context better than other kinds of devices and keep you present without picking up a mobile phone. Mos…
Inside Petal: Building the World’s First Petabit-Class Transoceanic Subsea Cable
Petal, the next step in Meta’s subsea innovation, will be the first subsea cable to deliver petabit capacity at transoceanic distances, connecting France and the United States over approximately 7,000 km (4,300 mi). Expe…
Open-Sourcing Rebalancer: A Generic, High-Performance Library for Solving Assignment Problems
We re open-sourcing Rebalancer, the assignment-problem solver that has been used to solve resource allocation problems throughout Meta for over nine years. Rebalancer separates several related concerns: how to specify an…
ZGateway: Learnings from Putting a Proxy in Front of ZippyDB
We’re introducing ZGateway, the proxy we are using to unify traffic through ZippyDB, Meta’s most widely-used key value store. As a bonus, it also enables admission control, load balancing, cross-region resilience, and ri…
An Organizational Second Brain: Building an AI That Learns From Experts
We’ve built an AI agent that acts as a secondary expert for a given domain, making deep specialist knowledge readily available and preserved for anyone in an organization to access, share, and build upon. This is not a t…
MetaRoCE: A New RDMA Transport Built for AI-Scale Ethernet
Training and serving frontier AI models depends on fast, reliable networks that move data between GPUs without wasting compute cycles. To meet this challenge at scale, Meta designed MetaRoCE – a clean-sheet RDMA transpor…
MTIA 300: Meta’s First Training Chip with Built-in NICs and Communication-Offloading Engines
MTIA 300 is the first of Meta’s family of in-house training and inference accelerators optimized for training ranking and recommendation models. We’re sharing how MTIA 300’s built-in NIC chiplets allow it to meet the com…
How We’re Building Scam Alert on WhatsApp With End-to-End Encryption and Verifiability Guarantees
WhatsApp is committed to helping people stay safe while protecting the privacy of their messages. As scam tactics evolve — from impersonation to social engineering to AI-generated lures — we re always evolving as well, s…
From User Sequences to Scaling Laws: A Multi-Stage Architecture for Meta’s Ads Ranking
Every day, Meta’s recommendation platforms handle billions of user interactions, generating rich temporal signals that capture individual preferences and intent across products, ads, and content. In our 2024 post on sequ…
GEM Training: How Meta Doubled the Efficiency of Its LLM-Scale Ads Foundation Model
Meta s Generative Ads Recommendation Model (GEM), the foundation model behind ads recommendations across Instagram and Facebook, now trains at LLM scale on several thousand of the latest-generation GPUs. This post goes i…
Exploring Hierarchical Interest Representation For Meta Ads Deep Funnel Optimization
Hierarchical Interest Representation is a research area for Meta Ads. We’re exploring an upstream representation layer over the universe of Ads entities – users, advertisers, products, services – learning unified embeddi…
Modernizing the Meta Ads Service With an Open-Source Kernel Scheduler
TL; DR At Meta s scale, a few milliseconds of latency degradation can have a significant negative impact on ads performance. When a Linux kernel upgrade risked regressing latency across Meta s ad serving fleet, we turned…
Meta’s AI Storage Blueprint at Scale
Over the past several years, model capabilities and training dataset sizes have experienced exponential growth. During the past year or so, the time between new-frontier-model releases has gone down from months to weeks.…
10 Years of Meta’s Commitment to Python
This year marks Meta s 10th consecutive year as a sponsor of the Python Software Foundation (PSF), the charitable organization dedicated to advancing, supporting, and protecting the open-source Python programming languag…
Privacy-Aware Infrastructure in the AI-Native Era: An Asset Classification Case Study
Privacy controls — systems that enforce retention, access, allowed-purpose, downstream-sharing, or anonymization policies — require a reliable understanding of data to function. Before such a control can operate effectiv…
How Meta Engineered Ultra-Narrow Batteries for AI Glasses
Smart glasses like the Ray-Ban Meta and Oakley Meta Vanguards need to pack enough energy to power features like cameras, speakers, AI workloads, and even a display. But it all has to fit into the glasses’ temple arms. So…
Adopting AV1 for Real-Time Communication (RTC) at Scale
Adopting AV1 for real-time communication at Meta has been a multi-year effort spanning codec selection, device eligibility, rate control, and error resilience. We’re sharing the technical and operational challenges while…
Lights Out, Systems On: Validating Instant Power Loss Readiness
We’re introducing Instantaneous PowerLoss Storm, a new testing paradigm within Meta’s infrastructure for handling and mitigating instant or zero-notice power loss in our data centers. We’re sharing: how we built readines…
SilverTorch: Index as Model — A New Retrieval Paradigm for Recommendation Systems
We’re introducing SilverTorch, a reimagining of recommendation systems that unifies all retrieval components for user generated content under a unified architecture. SilverTorch shows up to 23.7x higher throughput compar…
Reel Friends: Building Social Discovery that Scales to Billions
On its face the new Friend Bubbles feature looks simple enough. It highlights Reels your friends have watched and reacted to. But sometimes the features that seem the most straightforward require the deepest engineering…
Migrating Data Ingestion Systems at Meta Scale
Meta’s data ingestion system, which our engineering teams leverage for up-to-date snapshots of the social graph, has recently undergone a significant revamp to enhance its reliability at scale. Moving from our legacy sys…
Labyrinth 1.1: Making End-to-End Encrypted Backups Even More Reliable
We’re rolling out version 1.1 of Labyrinth, the encrypted storage system and protocol that secures messages and history on Messenger. Labyrinth 1.1 enhances the reliability of end-to-end encrypted backups with a new sub-…
How Meta Is Strengthening End-to-End Encrypted Backups
The HSM-based Backup Key Vault Meta s HSM-based Backup Key Vault provides the foundation for end-to-end encrypted backups for WhatsApp and Messenger. The system allows people to protect their backed-up message history wi…
Modernizing the Facebook Groups Search to Unlock the Power of Community Knowledge
We’ve fundamentally transformed Facebook Groups Search to help people more reliably discover, sort through, and validate community content that’s most relevant to them. We’ve adopted a new hybrid retrieval architecture a…
Capacity Efficiency at Meta: How Unified AI Agents Optimize Performance at Hyperscale
We re sharing insights into Meta s Capacity Efficiency Program, where we ve built an AI agent platform that helps automate finding and fixing performance issues throughout our infrastructure. By leveraging encoded domain…
Post-Quantum Cryptography Migration at Meta: Framework, Lessons, and Takeaways
We’re sharing lessons learned from Meta’s post-quantum cryptography (PQC) migration to help other organizations strengthen their resilience as industry transitions to post-quantum cryptography standards. We’re proposing…