# Snap

Welcome to our Snap Engineering 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.

## Alchemist: Agentic ML Experimentation

DevFeed: [Alchemist: Agentic ML Experimentation](<https://devfeed.tech/articles/alchemist-agentic-ml-experimentation-96304.md>)

Original publisher: [Read original article](<https://eng.snap.com/alchemist>)

Author: Wesley Chen

Published: 2026-10-22T21:07:38.510009Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Agentic AI](<https://devfeed.tech/topics/what-is-agentic-ai.md>), [Agent Harness](<https://devfeed.tech/topics/agent-harness.md>), [human-in-the-loop UX](<https://devfeed.tech/topics/human-in-the-loop-ux.md>), [Feature Engineering](<https://devfeed.tech/topics/feature-engineering.md>)

Tags: [agentic](<https://devfeed.tech/tags/agentic.md>), [agents](<https://devfeed.tech/tags/agents.md>), [cycle-time](<https://devfeed.tech/tags/cycle-time.md>), [cycles](<https://devfeed.tech/tags/cycles.md>), [experimentation](<https://devfeed.tech/tags/experimentation.md>), [feature-selection](<https://devfeed.tech/tags/feature-selection.md>), [harness](<https://devfeed.tech/tags/harness.md>), [human-in-the-loop](<https://devfeed.tech/tags/human-in-the-loop.md>), [ml](<https://devfeed.tech/tags/ml.md>), [snapchat](<https://devfeed.tech/tags/snapchat.md>)

### AI overview

Alchemist combines human-directed hypothesis selection with a deterministic harness to run ML experiments at Snapchat. Engineers set goals, curate candidate changes, and approve each plan; the harness validates plans, submits and monitors runs, collects results, and enforces budgets. It does not retry failures automatically: engineers review failure details and choose follow-up actions. In the feature-selection case study, a manual process removed 30% of features without regression in evaluation metrics after four weeks and 130 workflow submissions. The recipe reduced cycle time from four weeks to two. It uses permutation importance and learned stochastic feature gates. The stated results are from one Story ads model study; the text does not claim broader measured outcomes.

### Source excerpt

Alchemist brings human-in-the-loop agents and a deterministic harness to ML experimentation at Snapchat, reducing feature-selection cycles from four weeks to two.

## How Snap Built DS Agent as a Git Repository for Data-Science Work

DevFeed: [How Snap Built DS Agent as a Git Repository for Data-Science Work](<https://devfeed.tech/articles/page-56684.md>)

Original publisher: [Read original article](<https://eng.snap.com/ds_agent>)

Published: 2026-09-10T07:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Data Science](<https://devfeed.tech/topics/data-science.md>), [data](<https://devfeed.tech/topics/data.md>), [Git](<https://devfeed.tech/topics/git.md>), [Repository](<https://devfeed.tech/topics/repository.md>), [Orchestration](<https://devfeed.tech/topics/orchestration.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [data](<https://devfeed.tech/tags/data.md>), [data-science](<https://devfeed.tech/tags/data-science.md>), [git](<https://devfeed.tech/tags/git.md>), [orchestration](<https://devfeed.tech/tags/orchestration.md>)

### AI overview

Snap describes building DS Agent as a Git repository that supports data-science workflows inside a coding-agent environment, combining instructions, skills, domain knowledge, and guardrails.

### Source excerpt

DS Agent: Snap's AI Data Scientist Over the past year, some of the largest tech companies have written about their in-house data agents. OpenAI described a conversational agent over its warehouse with thousands of daily users. Meta shared an analytics agent adopted by most of its data scientists. Pinterest built the most-used internal agent at the company on top of its metadata platform.

## Building SPECS Experiences Faster

DevFeed: [Building SPECS Experiences Faster](<https://devfeed.tech/articles/building-specs-experiences-faster-56711.md>)

Original publisher: [Read original article](<https://eng.snap.com/specs_experiences>)

Published: 2026-08-20T07:00:00Z

Content type: tutorial

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [input](<https://devfeed.tech/topics/input.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [interface](<https://devfeed.tech/topics/interface.md>), [iteration](<https://devfeed.tech/topics/iteration.md>), [draggable](<https://devfeed.tech/topics/draggable.md>), [Hardware](<https://devfeed.tech/topics/hardware.md>), [Web](<https://devfeed.tech/topics/web.md>)

Tags: [agentic](<https://devfeed.tech/tags/agentic.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-coding](<https://devfeed.tech/tags/ai-coding.md>), [debugging](<https://devfeed.tech/tags/debugging.md>), [development](<https://devfeed.tech/tags/development.md>), [draggable](<https://devfeed.tech/tags/draggable.md>), [input](<https://devfeed.tech/tags/input.md>), [interface](<https://devfeed.tech/tags/interface.md>), [iteration](<https://devfeed.tech/tags/iteration.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [spatial](<https://devfeed.tech/tags/spatial.md>)

### AI overview

Snap describes a development workflow for SPECS spatial experiences using SIK for multimodal interaction, SPECS UI Kit for reusable spatial controls, and CLAD for AI-assisted scene assembly, debugging, testing, and iteration in Lens Studio.

### Source excerpt

Building SPECS Experiences Faster Building for SPECS is not like building a web page or a mobile app. The canvas is the world around the wearer, and the input model is hands, gaze, voice, mobile controls, and physical motion all at once. Get it wrong and the experience feels laggy, unreadable, or unstable the moment it's on someone's face.

## Casper: Snap's fleet of friendly AI teammates

DevFeed: [Casper: Snap's fleet of friendly AI teammates](<https://devfeed.tech/articles/casper-snap-s-fleet-of-friendly-ai-teammates-56677.md>)

Original publisher: [Read original article](<https://eng.snap.com/casper>)

Published: 2026-07-28T07:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [AI Agent](<https://devfeed.tech/topics/ai-agent.md>), [Software Engineering](<https://devfeed.tech/topics/software-engineering.md>), [AI-assisted coding](<https://devfeed.tech/topics/ai-assisted-coding.md>), [Code generation](<https://devfeed.tech/topics/code-generation.md>), [Code review](<https://devfeed.tech/topics/code-review.md>), [Pull Request](<https://devfeed.tech/topics/pull-request.md>), [Slack](<https://devfeed.tech/topics/slack.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>)

Tags: [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [code-generation](<https://devfeed.tech/tags/code-generation.md>), [code-review](<https://devfeed.tech/tags/code-review.md>), [coding](<https://devfeed.tech/tags/coding.md>), [coding-agent](<https://devfeed.tech/tags/coding-agent.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [development](<https://devfeed.tech/tags/development.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [human-in-the-loop](<https://devfeed.tech/tags/human-in-the-loop.md>), [identity-security](<https://devfeed.tech/tags/identity-security.md>), [pull-request](<https://devfeed.tech/tags/pull-request.md>)

### AI overview

Snap describes Casper, a fleet of autonomous AI agents that works through Slack and a web console to perform software engineering tasks, including coding, code review, data queries, project management, and pull request creation. The system keeps engineers involved in judgment decisions while handling repetitive work.

### Source excerpt

Casper: Snap's fleet of friendly AI teammates Software engineering is changing fast. Across the industry, the question is no longer whether AI can write production code, it's how much of the development lifecycle organizations are willing to offload to the agents. At Snap, we continue to increase the amount we count on agents for the work while keeping the human in the loop for judgement decisions.

## How GNSS Enables World-Scale Outdoor AR Tracking on Spectacles

DevFeed: [How GNSS Enables World-Scale Outdoor AR Tracking on Spectacles](<https://devfeed.tech/articles/gnss-tracking-for-specs-56712.md>)

Original publisher: [Read original article](<https://eng.snap.com/specs_gnss>)

Published: 2026-07-10T07:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [immersive web](<https://devfeed.tech/topics/immersive-web.md>), [3D](<https://devfeed.tech/topics/3d.md>), [data](<https://devfeed.tech/topics/data.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [augmented-reality](<https://devfeed.tech/tags/augmented-reality.md>), [device](<https://devfeed.tech/tags/device.md>), [gnss](<https://devfeed.tech/tags/gnss.md>), [orientation](<https://devfeed.tech/tags/orientation.md>), [precision](<https://devfeed.tech/tags/precision.md>), [sensors](<https://devfeed.tech/tags/sensors.md>)

### AI overview

This article explains how Spectacles combines GNSS with visual-inertial odometry to support geo-referenced six-degrees-of-freedom tracking for outdoor augmented reality. GNSS supplies absolute position and velocity, while local tracking provides high-frequency motion information but can drift over time.

### Source excerpt

World-scale Outdoor AR Tracking with GNSS Introduction Augmented reality (AR) has the power to transform the world around us, and its true potential is unlocked when digital content can be reliably anchored to specific, real-world locations. Imagine placing a digital object using geographic coordinates and seeing it there in 3D, or having navigational cues seamlessly overlaid onto your path, no matter when or where you look.

## Code Search: How Agents search across Snap's codebase

DevFeed: [Code Search: How Agents search across Snap's codebase](<https://devfeed.tech/articles/code-search-how-agents-search-across-snap-s-codebase-56678.md>)

Original publisher: [Read original article](<https://eng.snap.com/code_search>)

Published: 2026-07-01T07:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [code search](<https://devfeed.tech/topics/code-search.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [AI-assisted coding](<https://devfeed.tech/topics/ai-assisted-coding.md>), [Model Context Protocol](<https://devfeed.tech/topics/model-context-protocol.md>), [Command-line interface](<https://devfeed.tech/topics/cli.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [code-search](<https://devfeed.tech/tags/code-search.md>), [coding-agents](<https://devfeed.tech/tags/coding-agents.md>), [github](<https://devfeed.tech/tags/github.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [mcp](<https://devfeed.tech/tags/mcp.md>), [monorepo](<https://devfeed.tech/tags/monorepo.md>), [search](<https://devfeed.tech/tags/search.md>)

### AI overview

Snap replaced a single-machine regular-expression code search tool with a sharded, Zoekt-powered platform exposed through MCP. The system indexes thousands of repositories across multiple organizations and supports cross-repository discovery for engineers and AI coding agents.

### Source excerpt

Code Search: How Agents Search Across Snap's Codebase Snap's codebase spans thousands of repos across multiple organizations: millions of files and close to a terabyte of source. Until last year, our code search was a single VM regex tool that couldn't scale past one machine's memory, went dark during reindexing, and had no story for AI agents. Meanwhile, AI agents were multiplying fast, and the infrastructure they'd need to work across that codebase didn't exist yet.

## Measuring Spatial Intelligence in LLMs

DevFeed: [Measuring Spatial Intelligence in LLMs](<https://devfeed.tech/articles/measuring-spatial-intelligence-in-llms-56709.md>)

Original publisher: [Read original article](<https://eng.snap.com/spatial_intelligence>)

Published: 2026-06-16T07:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [Large Language Model](<https://devfeed.tech/topics/llm.md>), [LLM evaluation / benchmarking](<https://devfeed.tech/topics/llm-evaluation-benchmarking.md>), [Benchmark](<https://devfeed.tech/topics/benchmark.md>), [3D](<https://devfeed.tech/topics/3d.md>), [dataset](<https://devfeed.tech/topics/dataset.md>), [GitHub](<https://devfeed.tech/topics/github.md>), [Simulation](<https://devfeed.tech/topics/simulation.md>), [Computing](<https://devfeed.tech/topics/computing.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [computing](<https://devfeed.tech/tags/computing.md>), [dataset](<https://devfeed.tech/tags/dataset.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [games](<https://devfeed.tech/tags/games.md>), [github](<https://devfeed.tech/tags/github.md>), [leaderboard](<https://devfeed.tech/tags/leaderboard.md>), [llms](<https://devfeed.tech/tags/llms.md>), [measuring](<https://devfeed.tech/tags/measuring.md>), [models](<https://devfeed.tech/tags/models.md>), [simulation](<https://devfeed.tech/tags/simulation.md>), [spatial](<https://devfeed.tech/tags/spatial.md>)

### AI overview

The article introduces Spatial Benchmark, an open-source benchmark, dataset, and evaluation code for measuring how well language models reason about 3D space. It covers spatial tasks relevant to AR, XR, simulation, games, and embodied tools, and reports strong results alongside continuing gaps in engine-specific spatial reasoning.

### Source excerpt

Measuring Spatial Intelligence in LLMs Today we're open-sourcing Spatial Benchmark, a benchmark for measuring how well language models reason about 3D space. The benchmark, dataset, and evaluation code are available on GitHub, and you can browse the latest detailed results on our results summary page. It evaluates the kinds of spatial reasoning that matter in AR, XR, simulation, games, and embodied tools: placement, transforms, visibility, hierarchy, layout, and vector-based reasoning.

## CodePal: How Snap Built an AI Code Reviewer for the Age of AI-Written Code

DevFeed: [CodePal: How Snap Built an AI Code Reviewer for the Age of AI-Written Code](<https://devfeed.tech/articles/codepal-how-snap-built-an-ai-code-reviewer-for-the-age-of-ai-written-code-56679.md>)

Original publisher: [Read original article](<https://eng.snap.com/codepal>)

Published: 2026-06-05T07:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [ai code review](<https://devfeed.tech/topics/ai-code-review.md>), [AI-assisted coding](<https://devfeed.tech/topics/ai-assisted-coding.md>), [Code review](<https://devfeed.tech/topics/code-review.md>), [Software Engineering](<https://devfeed.tech/topics/software-engineering.md>), [Claude Code](<https://devfeed.tech/topics/claude-code.md>), [cursor](<https://devfeed.tech/topics/cursor.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-code-review](<https://devfeed.tech/tags/ai-code-review.md>), [ai-coding-tools](<https://devfeed.tech/tags/ai-coding-tools.md>), [claude-code](<https://devfeed.tech/tags/claude-code.md>), [code-review](<https://devfeed.tech/tags/code-review.md>), [cursor](<https://devfeed.tech/tags/cursor.md>)

### AI overview

Snap built CodePal, an internal AI-powered code review assistant, after AI coding tools increased pull-request volume and review bottlenecks. The system is designed to provide Snap-specific feedback across its engineering infrastructure and software development lifecycle.

### Source excerpt

CodePal: How Snap Built an AI Code Reviewer for the Age of AI-Written Code AI coding tools have fundamentally changed how software gets written. At Snap, engineers use tools like Cursor and Claude Code daily, and the impact on velocity has been real. Year-to-date, our merged (Pull Request) PR rate is up 60%. But there's a catch. When everyone can produce code faster, the bottleneck doesn't disappear, it shifts. And at Snap, it moved squarely onto the code review process.

## Snap Cloud: A Backend for Spectacles, Powered by Supabase

DevFeed: [Snap Cloud: A Backend for Spectacles, Powered by Supabase](<https://devfeed.tech/articles/snap-cloud-a-backend-for-spectacles-powered-by-supabase-56713.md>)

Original publisher: [Read original article](<https://eng.snap.com/spectacles_supabase>)

Published: 2026-04-17T07:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [Back end](<https://devfeed.tech/topics/backend.md>), [Entra ID](<https://devfeed.tech/topics/entra-id.md>), [PostgreSQL](<https://devfeed.tech/topics/postgresql.md>), [SQL](<https://devfeed.tech/topics/sql.md>), [Open Source](<https://devfeed.tech/topics/open-source.md>), [OAuth](<https://devfeed.tech/topics/oauth.md>), [Deno](<https://devfeed.tech/topics/deno.md>)

Tags: [authentication](<https://devfeed.tech/tags/authentication.md>), [backend](<https://devfeed.tech/tags/backend.md>), [databases](<https://devfeed.tech/tags/databases.md>), [developer-experience](<https://devfeed.tech/tags/developer-experience.md>), [multiplayer](<https://devfeed.tech/tags/multiplayer.md>), [oauth](<https://devfeed.tech/tags/oauth.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [supabase](<https://devfeed.tech/tags/supabase.md>)

### AI overview

Snap Cloud is an integrated backend platform for Spectacles developers, powered by Supabase. It provides databases, file storage, server-side logic, authentication, and realtime connectivity within Lens Studio, reducing the infrastructure setup required for AR experiences.

### Source excerpt

Snap Cloud: A Backend for Spectacles, Powered by Supabase Introduction Building compelling experiences for AR glasses often demands more than what runs on-device. Multiplayer games need shared state, multi-level games require persistent progress, fitness challenge Lenses need shared state and leaderboards. .

## Performance as a Core Product Feature

DevFeed: [Performance as a Core Product Feature](<https://devfeed.tech/articles/performance-as-a-core-product-feature-56698.md>)

Original publisher: [Read original article](<https://eng.snap.com/performance_as_a_feature>)

Published: 2026-03-19T07:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [App](<https://devfeed.tech/topics/app.md>), [Mobile](<https://devfeed.tech/topics/mobile.md>), [Latency](<https://devfeed.tech/topics/latency.md>), [tracing](<https://devfeed.tech/topics/tracing.md>), [Monitoring](<https://devfeed.tech/topics/monitoring.md>), [Users](<https://devfeed.tech/topics/users.md>), [User Experience](<https://devfeed.tech/topics/user-experience.md>), [Grafana](<https://devfeed.tech/topics/grafana.md>)

Tags: [app](<https://devfeed.tech/tags/app.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [mobile-performance](<https://devfeed.tech/tags/mobile-performance.md>), [monitoring](<https://devfeed.tech/tags/monitoring.md>), [performance](<https://devfeed.tech/tags/performance.md>), [product](<https://devfeed.tech/tags/product.md>), [reliability](<https://devfeed.tech/tags/reliability.md>), [user-experience](<https://devfeed.tech/tags/user-experience.md>)

### AI overview

Snapchat treats performance as a core product feature and focuses on delivering a fast, reliable camera-opening experience across devices and network conditions. The article explains how the team uses p90 tail-latency monitoring, segmented analysis, rollout gates, and a custom production tracing system to identify performance problems while minimizing runtime overhead.

### Source excerpt

Performance as a Core Product Feature At Snapchat, we don't just see performance as a requirement; we treat it as a core product feature. For millions of users, the app is the camera. That critical "open-to-camera" flow needs to be instant and reliable, every single time. But how do you ensure a quality experience for all users, not just the median? As our app and team grew, we needed a way to find and fix performance issues without impacting the user experience..

## How Snap Evolved Flowrida from a Monolithic Airflow Cluster to a Resilient Multi-Cluster Platform

DevFeed: [How Snap Evolved Flowrida from a Monolithic Airflow Cluster to a Resilient Multi-Cluster Platform](<https://devfeed.tech/articles/from-monolith-to-multi-cluster-56687.md>)

Original publisher: [Read original article](<https://eng.snap.com/flowrida>)

Published: 2026-03-17T07:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [Orchestration](<https://devfeed.tech/topics/orchestration.md>), [airflow](<https://devfeed.tech/topics/airflow.md>), [Data pipelines](<https://devfeed.tech/topics/data-pipelines.md>), [Scalability](<https://devfeed.tech/topics/scalability.md>), [data](<https://devfeed.tech/topics/data.md>), [data-processing](<https://devfeed.tech/topics/data-processing.md>), [Availability](<https://devfeed.tech/topics/availability.md>), [Kubernetes](<https://devfeed.tech/topics/kubernetes.md>), [Disaster Recovery](<https://devfeed.tech/topics/disaster-recovery.md>), [Open Source](<https://devfeed.tech/topics/open-source.md>), [Database](<https://devfeed.tech/topics/database.md>), [A/B Testing](<https://devfeed.tech/topics/a-b-testing.md>)

Tags: [apache-airflow](<https://devfeed.tech/tags/apache-airflow.md>), [clusters](<https://devfeed.tech/tags/clusters.md>), [data-pipelines](<https://devfeed.tech/tags/data-pipelines.md>), [database](<https://devfeed.tech/tags/database.md>), [disaster-recovery](<https://devfeed.tech/tags/disaster-recovery.md>), [high-availability](<https://devfeed.tech/tags/high-availability.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [multicluster](<https://devfeed.tech/tags/multicluster.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [orchestration](<https://devfeed.tech/tags/orchestration.md>), [scale](<https://devfeed.tech/tags/scale.md>)

### AI overview

Snap describes how it evolved Flowrida, an Apache Airflow-based workflow orchestration platform, from a monolithic Kubernetes cluster to a multicluster architecture. The changes addressed availability, disaster recovery, deployment blast radius, recovery time, and scalability challenges as data-processing demands grew.

### Source excerpt

From Monolith to Multi-Cluster: Snap's Journey to a Resilient Data Orchestration Platform Flowrida, Snap's workflow orchestration tool, is built on Apache Airflow and is critical for managing data at scale. It supports over 1,000 users across more than 170 teams, handling 3,300 active data pipelines and over 180,000 task instances daily. A significant limitation of open-source Apache Airflow is its lack of multi-cluster resiliency.

## Inside Snap's Experimentation Platform: Leveraging NVIDIA GPUs for Accelerated Data Processing

DevFeed: [Inside Snap's Experimentation Platform: Leveraging NVIDIA GPUs for Accelerated Data Processing](<https://devfeed.tech/articles/inside-snap-s-experimentation-platform-leveraging-nvidia-gpus-for-accelerated-data-processing-56706.md>)

Original publisher: [Read original article](<https://eng.snap.com/snap-nvidia-gcp>)

Published: 2026-03-16T07:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [experiments](<https://devfeed.tech/topics/experiments.md>), [Apache Spark](<https://devfeed.tech/topics/spark.md>), [GPU](<https://devfeed.tech/topics/gpu.md>), [Data pipelines](<https://devfeed.tech/topics/data-pipelines.md>), [RAPIDS](<https://devfeed.tech/topics/rapids.md>), [Google Cloud Platform (GCP)](<https://devfeed.tech/topics/google-cloud.md>), [Nvidia](<https://devfeed.tech/topics/nvidia.md>), [Cloud](<https://devfeed.tech/topics/cloud.md>)

Tags: [apache-spark](<https://devfeed.tech/tags/apache-spark.md>), [cost](<https://devfeed.tech/tags/cost.md>), [cost-savings](<https://devfeed.tech/tags/cost-savings.md>), [data-pipelines](<https://devfeed.tech/tags/data-pipelines.md>), [experimentation](<https://devfeed.tech/tags/experimentation.md>), [google-cloud](<https://devfeed.tech/tags/google-cloud.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [gpu-accelerated](<https://devfeed.tech/tags/gpu-accelerated.md>)

### AI overview

Snap describes migrating large-scale experimentation data pipelines to GPU-accelerated Apache Spark on Google Cloud. The post covers reliability, operational complexity, shared GPU infrastructure, and reported cost savings.

### Source excerpt

Inside Snap's Experimentation Platform: Leveraging NVIDIA GPUs for Accelerated Data Processing Migrating Snapchat's AB Pipelines to GPU-Accelerated Spark At Snap, experimentation sits at the core of how we build products. Nearly every product decision is informed by controlled experiments, and the infrastructure that powers those experiments needs to be reliable, scalable, and fast.

## Building the Spatial Interaction and Interface Frameworks for Specs

DevFeed: [Building the Spatial Interaction and Interface Frameworks for Specs](<https://devfeed.tech/articles/building-the-spatial-interaction-and-interface-frameworks-for-specs-56710.md>)

Original publisher: [Read original article](<https://eng.snap.com/spatial_interaction>)

Published: 2026-03-13T07:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [Framework](<https://devfeed.tech/topics/framework.md>), [interface](<https://devfeed.tech/topics/interface.md>), [Computing](<https://devfeed.tech/topics/computing.md>), [Code](<https://devfeed.tech/topics/code.md>), [Hardware](<https://devfeed.tech/topics/hardware.md>)

Tags: [computing](<https://devfeed.tech/tags/computing.md>), [design-principles](<https://devfeed.tech/tags/design-principles.md>), [developers](<https://devfeed.tech/tags/developers.md>), [development](<https://devfeed.tech/tags/development.md>), [frameworks](<https://devfeed.tech/tags/frameworks.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [interaction](<https://devfeed.tech/tags/interaction.md>), [interface](<https://devfeed.tech/tags/interface.md>), [source-code](<https://devfeed.tech/tags/source-code.md>), [ui](<https://devfeed.tech/tags/ui.md>)

### AI overview

Snap describes how it built the Spectacles Interaction Kit and Spectacles UI Kit, spatial frameworks for hand-, voice-, and controller-based interactions and interfaces in augmented reality. The post covers their design principles, source-code extensibility, gesture handling, targeting accuracy, wearable performance, and UI component architecture.

### Source excerpt

Building the Spatial Interaction and Interface Frameworks for Specs An inside look at our process for building the interaction and interface frameworks for Specs: Spectacles Interaction Kit (SIK) and Spectacles UI Kit (UIKit). For over a decade, Snap has been building toward a new computing paradigm. One that keeps you present in the real world instead of pulling you into a screen.

## A Declarative Standard for AI Agents

DevFeed: [A Declarative Standard for AI Agents](<https://devfeed.tech/articles/a-declarative-standard-for-ai-agents-56673.md>)

Original publisher: [Read original article](<https://eng.snap.com/agent-format>)

Published: 2026-03-11T07:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [standard](<https://devfeed.tech/topics/standard.md>), [LangChain](<https://devfeed.tech/topics/langchain.md>), [google-adk](<https://devfeed.tech/topics/google-adk.md>), [Framework](<https://devfeed.tech/topics/framework.md>), [Python](<https://devfeed.tech/topics/python.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [declarative](<https://devfeed.tech/tags/declarative.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [framework](<https://devfeed.tech/tags/framework.md>), [google-adk](<https://devfeed.tech/tags/google-adk.md>), [governance](<https://devfeed.tech/tags/governance.md>), [guardrails](<https://devfeed.tech/tags/guardrails.md>), [implementing](<https://devfeed.tech/tags/implementing.md>), [langchain](<https://devfeed.tech/tags/langchain.md>), [python](<https://devfeed.tech/tags/python.md>), [standard](<https://devfeed.tech/tags/standard.md>)

### AI overview

Snap describes a proposed declarative standard for AI agents intended to reduce duplicated work across teams using different frameworks. The approach separates agent definitions from the underlying runtime, drawing inspiration from Kubernetes to improve portability, shared tooling, and governance.

### Source excerpt

Agent Format: A Declarative Standard for AI Agents Ten months ago, three engineering teams at Snap were independently building AI agents. One team used LangChain. Another picked Google ADK. A third wrote custom Python scripts. All three agents did something similar and all three worked well. But as the number of agents grew, we spotted an opportunity. Each team was solving the same problems in isolation: wiring up tools, implementing retry logic, tracking token budgets, building output parsers.

## Spectacles - EyeConnect

DevFeed: [Spectacles - EyeConnect](<https://devfeed.tech/articles/spectacles-eyeconnect-56686.md>)

Original publisher: [Read original article](<https://eng.snap.com/eyeconnect>)

Published: 2026-01-27T08:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [immersive web](<https://devfeed.tech/topics/immersive-web.md>), [3D](<https://devfeed.tech/topics/3d.md>), [Algorithms](<https://devfeed.tech/topics/algorithms.md>), [Bluetooth](<https://devfeed.tech/topics/bluetooth.md>), [webcam](<https://devfeed.tech/topics/webcam.md>), [User interface design](<https://devfeed.tech/topics/ui-design.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [algorithm](<https://devfeed.tech/tags/algorithm.md>), [augmented-reality](<https://devfeed.tech/tags/augmented-reality.md>), [bluetooth](<https://devfeed.tech/tags/bluetooth.md>), [camera](<https://devfeed.tech/tags/camera.md>), [latency](<https://devfeed.tech/tags/latency.md>), [privacy](<https://devfeed.tech/tags/privacy.md>), [slam](<https://devfeed.tech/tags/slam.md>), [tracking](<https://devfeed.tech/tags/tracking.md>), [user-interface](<https://devfeed.tech/tags/user-interface.md>)

### AI overview

Snap describes EyeConnect, a Spectacles feature for joining shared augmented-reality sessions by having two users look at each other. It aligns users' virtual worlds using tracked device motion, head poses, and Spectacles detections exchanged over Bluetooth, while deleting the alignment data after use.

### Source excerpt

Spectacles - EyeConnect Shared Augmented Reality experiences are part of Spectacles' mission to connect humans in the real world. In a shared experience, each user can see and interact with the digital content from their own perspective as if it were real. Traditional methods require all users to scan the environment or fiducials placed in the environment; this can be cumbersome in certain situations.

## Universal User Modeling (UUM): A Foundation Model for User Understanding at Snapchat

DevFeed: [Universal User Modeling (UUM): A Foundation Model for User Understanding at Snapchat](<https://devfeed.tech/articles/universal-user-modeling-uum-a-foundation-model-for-user-understanding-at-snapchat-56719.md>)

Original publisher: [Read original article](<https://eng.snap.com/universal_user_modeling>)

Published: 2025-12-12T07:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [recommendation systems](<https://devfeed.tech/topics/recommendation-systems.md>), [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [personalization](<https://devfeed.tech/topics/personalization.md>), [data-processing](<https://devfeed.tech/topics/data-processing.md>), [Feature Engineering](<https://devfeed.tech/topics/feature-engineering.md>), [Logging](<https://devfeed.tech/topics/logging.md>), [dataset](<https://devfeed.tech/topics/dataset.md>), [data](<https://devfeed.tech/topics/data.md>)

Tags: [data](<https://devfeed.tech/tags/data.md>), [data-processing](<https://devfeed.tech/tags/data-processing.md>), [dataset](<https://devfeed.tech/tags/dataset.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [feature-engineering](<https://devfeed.tech/tags/feature-engineering.md>), [foundation](<https://devfeed.tech/tags/foundation.md>), [logging](<https://devfeed.tech/tags/logging.md>), [ml](<https://devfeed.tech/tags/ml.md>), [model](<https://devfeed.tech/tags/model.md>), [personalization](<https://devfeed.tech/tags/personalization.md>), [user](<https://devfeed.tech/tags/user.md>)

### AI overview

Snap describes Universal User Modeling, a cross-surface foundation model that creates reusable long-term user embeddings from behavior across Snapchat products. The model supplements surface-specific models and supports personalization across use cases including Stories, Ads, Spotlight, notifications, Lens, and content search.

### Source excerpt

Prism is a Spark-powered platform at Snap that simplifies ML data processing, enhancing speed & consistency.

## Building a Spark-Powered Platform for ML Data Needs at Snap

DevFeed: [Building a Spark-Powered Platform for ML Data Needs at Snap](<https://devfeed.tech/articles/building-a-spark-powered-platform-for-ml-data-needs-at-snap-56700.md>)

Original publisher: [Read original article](<https://eng.snap.com/prism>)

Published: 2025-09-05T07:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [Apache Spark](<https://devfeed.tech/topics/spark.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Data Infrastructure](<https://devfeed.tech/topics/data-infrastructure.md>), [Data pipelines](<https://devfeed.tech/topics/data-pipelines.md>), [data-platforms](<https://devfeed.tech/topics/data-platforms.md>), [data-processing](<https://devfeed.tech/topics/data-processing.md>)

Tags: [data](<https://devfeed.tech/tags/data.md>), [data-infrastructure](<https://devfeed.tech/tags/data-infrastructure.md>), [data-pipelines](<https://devfeed.tech/tags/data-pipelines.md>), [data-platform](<https://devfeed.tech/tags/data-platform.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [ml](<https://devfeed.tech/tags/ml.md>), [parquet](<https://devfeed.tech/tags/parquet.md>), [spark](<https://devfeed.tech/tags/spark.md>)

### AI overview

Snap describes Prism, a Spark-powered platform designed to address the data-processing needs of machine-learning teams. The platform aims to support large-scale experimentation, flexible development workflows, production stability, and diverse data formats.

### Source excerpt

Prism is a Spark-powered platform at Snap that simplifies ML data processing, enhancing speed & consistency.

## Snap Ads Understanding: From Pixels to Words

DevFeed: [Snap Ads Understanding: From Pixels to Words](<https://devfeed.tech/articles/snap-ads-understanding-from-pixels-to-words-56704.md>)

Original publisher: [Read original article](<https://eng.snap.com/snap-ads-understanding>)

Published: 2025-04-23T20:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [recommendation systems](<https://devfeed.tech/topics/recommendation-systems.md>), [systems](<https://devfeed.tech/topics/systems.md>)

Tags: [embedding-model](<https://devfeed.tech/tags/embedding-model.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [moderation](<https://devfeed.tech/tags/moderation.md>), [platform](<https://devfeed.tech/tags/platform.md>), [policy](<https://devfeed.tech/tags/policy.md>), [ranking](<https://devfeed.tech/tags/ranking.md>), [safety](<https://devfeed.tech/tags/safety.md>), [scale](<https://devfeed.tech/tags/scale.md>), [snap](<https://devfeed.tech/tags/snap.md>)

### AI overview

Snap describes a technique for converting advertising creatives, including videos and images, into rich, structured text. The resulting concepts support scalable ad moderation, policy enforcement, relevance and ranking models, safety analysis, trend detection, and creative analysis. The article explains limitations of embedding-only approaches, including lost visual context, poor interpretability, and difficulty adding explicit labels or querying specific attributes.

### Source excerpt

Our advertising platform at Snap receives a large volume of new ads every day. To maintain a high-quality and relevant ad ecosystem, we need to review these creatives at scale, to ensure they align with our policies, are safe for our community, and are shown to the right Snapchatters.

## Introducing Bento, Snap's ML Platform

DevFeed: [Introducing Bento, Snap's ML Platform](<https://devfeed.tech/articles/introducing-bento-snap-s-ml-platform-56691.md>)

Original publisher: [Read original article](<https://eng.snap.com/introducing-bento>)

Author: Snap Platform Engineering Team

Published: 2025-01-28T20:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [MLOps](<https://devfeed.tech/topics/mlops.md>), [model-deployment](<https://devfeed.tech/topics/model-deployment.md>), [feature-store](<https://devfeed.tech/topics/feature-store.md>), [data](<https://devfeed.tech/topics/data.md>), [big-data](<https://devfeed.tech/topics/big-data.md>), [Open Source](<https://devfeed.tech/topics/open-source.md>)

Tags: [architecture](<https://devfeed.tech/tags/architecture.md>), [array](<https://devfeed.tech/tags/array.md>), [data](<https://devfeed.tech/tags/data.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [feature-store](<https://devfeed.tech/tags/feature-store.md>), [ml](<https://devfeed.tech/tags/ml.md>), [ml-engineering](<https://devfeed.tech/tags/ml-engineering.md>), [ml-ops](<https://devfeed.tech/tags/ml-ops.md>), [ml-platform](<https://devfeed.tech/tags/ml-platform.md>), [model-deployment](<https://devfeed.tech/tags/model-deployment.md>), [model-training](<https://devfeed.tech/tags/model-training.md>), [models](<https://devfeed.tech/tags/models.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [platform](<https://devfeed.tech/tags/platform.md>), [scale](<https://devfeed.tech/tags/scale.md>)

### AI overview

Snap's Bento is an ML platform for managing server-side machine-learning workloads at scale. The article explains its architecture, end-to-end development lifecycle, feature and training data pipelines, model experimentation, deployment, and the operational challenges of running real-time ML applications in production.

### Source excerpt

Building ML at scale: Discover how Snap's Bento ML Platform powers personalized user experiences through cutting-edge ML Ops

## Embedding-based Retrieval with Two-Tower Models in Spotlight

DevFeed: [Embedding-based Retrieval with Two-Tower Models in Spotlight](<https://devfeed.tech/articles/embedding-based-retrieval-with-two-tower-models-in-spotlight-56685.md>)

Original publisher: [Read original article](<https://eng.snap.com/embedding-based-retrieval>)

Author: Snap Discover Team

Published: 2023-06-06T07:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [recommendation systems](<https://devfeed.tech/topics/recommendation-systems.md>), [Representation learning](<https://devfeed.tech/topics/representation-learning.md>), [information retrieval](<https://devfeed.tech/topics/information-retrieval.md>), [Latency](<https://devfeed.tech/topics/latency.md>), [personalization](<https://devfeed.tech/topics/personalization.md>), [real-time](<https://devfeed.tech/topics/real-time.md>)

Tags: [embedding](<https://devfeed.tech/tags/embedding.md>), [latency](<https://devfeed.tech/tags/latency.md>), [models](<https://devfeed.tech/tags/models.md>), [personalization](<https://devfeed.tech/tags/personalization.md>), [ranking](<https://devfeed.tech/tags/ranking.md>), [real-time](<https://devfeed.tech/tags/real-time.md>), [recommendation-system](<https://devfeed.tech/tags/recommendation-system.md>), [retrieval](<https://devfeed.tech/tags/retrieval.md>), [scale](<https://devfeed.tech/tags/scale.md>), [two-tower-model](<https://devfeed.tech/tags/two-tower-model.md>)

### AI overview

Snap describes the challenges and techniques behind large-scale embedding-based retrieval for Spotlight, its short-video platform. The system uses embeddings and a two-tower model to retrieve personalized video candidates from a large content pool while meeting real-time serving latency requirements.

### Source excerpt

At Snap, we want to empower you to learn about the world and have fun with other Snapchatters. On Spotlight, Snap's short-video platform, we personalize what you see based on your interests and previous watch history. This blog discusses the challenges we faced and techniques we applied for large-scale embedding-based retrieval, which selects the stories that users may like from a large candidate pool in real time.

## Differential Privacy at Snapchat

DevFeed: [Differential Privacy at Snapchat](<https://devfeed.tech/articles/differential-privacy-at-snapchat-56682.md>)

Original publisher: [Read original article](<https://eng.snap.com/differential-privacy-at-snap>)

Author: Snap Security Team

Published: 2022-12-15T07:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [Differential Privacy](<https://devfeed.tech/topics/differential-privacy.md>), [data](<https://devfeed.tech/topics/data.md>), [standard](<https://devfeed.tech/topics/standard.md>)

Tags: [article](<https://devfeed.tech/tags/article.md>), [data](<https://devfeed.tech/tags/data.md>), [differential-privacy](<https://devfeed.tech/tags/differential-privacy.md>), [privacy](<https://devfeed.tech/tags/privacy.md>), [product](<https://devfeed.tech/tags/product.md>), [standard](<https://devfeed.tech/tags/standard.md>)

### AI overview

Snap explains how it uses differential privacy to protect user data in product features such as friend recommendations and Maps. The article describes differential privacy as a technique that adds calibrated noise to data computations so individual records are difficult to infer from query outputs.

### Source excerpt

At Snap, privacy is a central part of our product philosophy. We want Snapchatters to feel safe and secure interacting with their friends and exploring the world around them without worrying about the privacy of their data. To achieve this goal one of the tools we use is Differential Privacy (DP), the de facto privacy standard in both industry and academia. In this article, first we summarize some of the privacy concerns that Snap and the industry as a whole face. Then we examine how Snap uses DP to address those concerns regarding friend recommendations and Maps features.

## AR-Enabled Catalog

DevFeed: [AR-Enabled Catalog](<https://devfeed.tech/articles/ar-enabled-catalog-56675.md>)

Original publisher: [Read original article](<https://eng.snap.com/ar-enabled-catalogs>)

Author: Catalog Platform Team

Published: 2022-11-11T07:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [data](<https://devfeed.tech/topics/data.md>), [Orchestration](<https://devfeed.tech/topics/orchestration.md>), [processing](<https://devfeed.tech/topics/processing.md>), [service](<https://devfeed.tech/topics/service.md>), [3D](<https://devfeed.tech/topics/3d.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [data](<https://devfeed.tech/tags/data.md>), [ecommerce](<https://devfeed.tech/tags/ecommerce.md>), [orchestration](<https://devfeed.tech/tags/orchestration.md>), [processing](<https://devfeed.tech/tags/processing.md>), [service](<https://devfeed.tech/tags/service.md>)

### AI overview

Snap describes how its catalog platform combines primary product-data feeds with supplemental feeds containing AR asset links, allowing advertisers to support AR-enabled products without modifying existing feeds.

### Source excerpt

Behind the lens: How supplemental feeds help advertisers bring the AR experience to Snap ads and reach millions of Snapchatters.

## Speed Up Feature Engineering for Recommendations

DevFeed: [Speed Up Feature Engineering for Recommendations](<https://devfeed.tech/articles/speed-up-feature-engineering-for-recommendations-56714.md>)

Original publisher: [Read original article](<https://eng.snap.com/speed-up-feature-engineering>)

Author: Snap Feature Platform team

Published: 2022-09-29T19:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [Feature Engineering](<https://devfeed.tech/topics/feature-engineering.md>), [recommendation systems](<https://devfeed.tech/topics/recommendation-systems.md>), [data](<https://devfeed.tech/topics/data.md>), [Grafana](<https://devfeed.tech/topics/grafana.md>)

Tags: [engineering](<https://devfeed.tech/tags/engineering.md>), [feature-engineering](<https://devfeed.tech/tags/feature-engineering.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [logging](<https://devfeed.tech/tags/logging.md>), [ml](<https://devfeed.tech/tags/ml.md>), [platform](<https://devfeed.tech/tags/platform.md>), [recommendations](<https://devfeed.tech/tags/recommendations.md>), [user-experience](<https://devfeed.tech/tags/user-experience.md>)

### AI overview

Snap describes Robusta, an in-house feature platform designed to streamline feature engineering and accelerate machine-learning iteration for recommendation systems. The platform addresses slow feature extraction, risky changes to serving components, duplicated infrastructure, and limited feature sharing.

### Source excerpt

At Snap, we have a strong focus on user experience and have built many ML applications to surface the right content to our users. We present Robusta: our in-house feature platform, built to streamline and accelerate ML iteration.

## Synthetic Data for Machine Learning

DevFeed: [Synthetic Data for Machine Learning](<https://devfeed.tech/articles/synthetic-data-for-machine-learning-56716.md>)

Original publisher: [Read original article](<https://eng.snap.com/synthetic-data-for-machine-learning>)

Author: Cameos Team

Published: 2022-09-01T08:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [synthetic-data](<https://devfeed.tech/topics/synthetic-data.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [AI Development](<https://devfeed.tech/topics/ai-development.md>), [dataset](<https://devfeed.tech/topics/dataset.md>), [data-processing](<https://devfeed.tech/topics/data-processing.md>), [3D](<https://devfeed.tech/topics/3d.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [data](<https://devfeed.tech/tags/data.md>), [dataset](<https://devfeed.tech/tags/dataset.md>), [datasets](<https://devfeed.tech/tags/datasets.md>), [developers](<https://devfeed.tech/tags/developers.md>), [development](<https://devfeed.tech/tags/development.md>), [generator](<https://devfeed.tech/tags/generator.md>), [labeling](<https://devfeed.tech/tags/labeling.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [models](<https://devfeed.tech/tags/models.md>), [synthetic](<https://devfeed.tech/tags/synthetic.md>), [synthetic-data](<https://devfeed.tech/tags/synthetic-data.md>)

### AI overview

Snap describes using synthetic data to accelerate machine-learning model development. The approach addresses dataset bias, missing corner cases, expensive or difficult labeling, licensing constraints, and varied data requirements through a configurable synthetic-data pipeline and generator built from reusable components.

### Source excerpt

At Snap we develop a large number of products and solutions powered by ML. Modern ML models require tons of high-quality data to be collected, labeled and processed. All these processes may frequently take a long time and lots of resources. In this post we talk about how we leverage synthetic data to boost the development of our ML models.

[Next page](<https://devfeed.tech/sources/snap.md?cursor=WyIyMDIyLTA5LTAxVDA4OjAwOjAwKzAwOjAwIiwgIjFhMWY4ZDJkLWJhYWMtNGIxNy04MzE3LWE5MGZkODg5YTgzMyJd>)