Techlist.io - Korean Tech Blog Curator

figmaOct 27, 20253 min readCurated summary

Schema 2025: Design Systems For A New Era | Figma Blog

Figma’s Schema 2025 announcements present design systems as living infrastructure for an AI-driven product development process. They are evolving beyond static rules into a shared language connecting design, code, people, and AI. The updates focus on scaling across brands, enabling flexibility without sacrificing consistency, and improving the connection between design intent and implementation. ## Design systems for the AI era - As product, design, and engineering roles increasingly overlap, more people and AI tools contribute to product development. - Design systems can provide the common language needed to maintain consistency across these contributors. - Figma aims to help teams preserve quality and design intent while allowing ideas to evolve across products and platforms. - The announced features emphasize: - Power balanced with flexibility - Better connections between design and code - Broader participation in the design process ## Extended collections for multi-brand systems - Traditional variables work well for straightforward theming but can be limiting for organizations with multiple products and distinct brands. - Extended collections allow teams to create a white-labeled version of a core design system. - Individual teams can add, publish, and reuse their own themes while remaining connected to the parent system. - Extensions automatically inherit updates such as new variables or color changes. - Explicitly overridden values remain customized, allowing local flexibility without losing synchronization. - Extended collections are expected to become available in November. ## Slots for flexible components - Conventional Figma components restrict how designers can modify instances. - For example, dropdown components traditionally required hidden list items or detaching the component to add new content. - Slots will allow designers to insert their own layers inside component instances without breaking the connection to the design system. - Component authors can specify which types of instances a slot accepts. - This improves usability while preserving design-system compliance. - Slots are available through an early-access program. ## Check designs for better token usage - Developers often need clarification about which exact design token corresponds to a raw value in a design. - The Check designs linter identifies elements that should align with the design system, including variables. - Figma’s custom model suggests the appropriate variable for each context. - Designers can review suggestions before applying them and then hand off more reliable designs to development. - Early access is available to organizations and Enterprise full-seat plans. Figma’s direction is to make design systems more adaptable and intelligent: centralized enough to preserve consistency, but flexible enough for multiple brands, contributors, and use cases. Teams should look toward systems that can actively guide design and implementation rather than merely document standards.

Read original

Figma AI Design Systems Design Tokens+1

googleOct 26, 2025Original article

How we are building the personal health coach (opens in new tab)

Google is leveraging Gemini models to create a proactive, adaptive personal health coach designed to bridge the gap between fragmented health data and actionable wellness guidance. By integrating physiological metrics with behavioral science, the system provides tailored insights and sustainable habit-building plans through a sophisticated multi-agent AI architecture. This initiative, currently in public preview for Fitbit Premium users, represents a transition toward data-driven, expert-validated health coaching that evolves dynamically with an individual's progress. ## Architecting a Multi-Agent Health Coach The system utilizes a complex multi-agent framework to coordinate different specialized AI sub-agents, ensuring that health recommendations are holistic and contextually aware. * **Conversational Agent:** Manages multi-turn interactions, understands user intent, and orchestrates the other agents while gathering necessary context for response generation. * **Data Science Agent:** Employs code-generation capabilities to iteratively fetch, analyze, and summarize physiological time-series data, such as sleep patterns and workout intensity. * **Domain Expert Agent:** Analyzes user data through the lens of specific fields like fitness or nutrition to generate and adapt personalized plans based on changing user context. * **Numerical Reasoning:** The coach performs sophisticated reasoning on health metrics, comparing current data against personal baselines and population-level statistics using capabilities derived from PH-LLM research. ## Ensuring Reliability via the SHARP Framework To move beyond general-purpose AI capabilities, the system is grounded in established coaching frameworks and subjected to rigorous technical and clinical validation. * **SHARP Evaluation:** The model is continuously assessed across five dimensions: Safety, Helpfulness, Accuracy, Relevance, and Personalization. * **Human-in-the-Loop Validation:** The development process involved over 1 million human annotations and 100,000 hours of evaluation by specialists in fields such as cardiology, endocrinology, and behavioral science. * **Expert Oversight:** Google convened a Consumer Health Advisory Panel and collaborated with professional fitness coaches to ensure the AI's recommendations align with real-world professional standards. * **Scientific Grounding:** The coach utilizes novel methods to foster consensus in nuanced health areas, ensuring that wellness recommendations remain scientifically accurate through the use of scaled "autoraters." Eligible Fitbit Premium users on Android in the US can now opt into the public preview to provide feedback on these personalized insights. As the tool evolves through iterative design and user research, it aims to provide a seamless connection between raw health metrics and sustainable lifestyle changes.

Code Generation Multi Agent Systems Agent Orchestration Ai Evaluation+4

netflixOct 25, 2025Original article

Post-Training Generative Recommenders with Advantage-Weighted Supervised Finetuning | by Netflix Technology Blog | Netflix TechBlog (opens in new tab)

Netflix is evolving its recommendation systems by moving beyond simple behavior imitation toward generative recommenders that better align with true user preferences. While generative models like HSTU and OneRec effectively capture sequential user patterns, they often struggle to distinguish between habitual clicks and genuine satisfaction. To bridge this gap, Netflix developed Advantage-Weighted Supervised Fine-tuning (A-SFT), a post-training method that leverages noisy reward signals to refine model performance without the need for complex counterfactual data. ### The Shift to Generative Recommenders * Modern generative recommenders (GRs), such as HSTU and OneRec, utilize transformer architectures to treat recommendation as a sequential transduction task. * The models are typically trained using next-item prediction, where the system learns to imitate the chronological sequence of a user’s activities. * A significant drawback of this "behavior cloning" approach is that it captures external trends and noise rather than long-term user satisfaction, potentially recommending content the user finished but did not actually enjoy. ### Barriers to Reinforcement Learning in RecSys * Traditional post-training methods used in Large Language Models, such as Proximal Policy Optimization (PPO) or Direct Preference Optimization (DPO), require counterfactual feedback that is difficult to obtain in recommendation contexts. * Because user sequences span weeks or years, it is impractical to generate and test hypothetical, counterfactual experiences for real-time user validation. * Reward signals in recommendation systems are inherently noisy; for instance, high watch time might indicate interest, but it can also be a result of external circumstances, making it an unreliable metric for optimization. ### Advantage-Weighted Supervised Fine-tuning (A-SFT) * A-SFT is a hybrid approach that sits between offline reinforcement learning and standard supervised fine-tuning. * The algorithm incorporates an advantage function to weight training examples, allowing the model to prioritize actions that lead to higher rewards while filtering out noise from the reward model. * This method is specifically designed to handle high-variance reward signals, using them as directional guides rather than absolute truth, which prevents the model from over-exploiting inaccurate data. * Benchmarks against other representative methods show that A-SFT achieves superior alignment between the generative recommendation policy and the underlying reward model. For organizations managing large-scale recommendation engines, A-SFT offers a practical path to implementing post-training improvements. By focusing on advantage-weighted signals, developers can improve recommendation quality using existing implicit feedback—like watch time and clicks—without the infrastructure hurdles of online reinforcement learning.

AI Machine Learning Transformer Recommendation Systems+4

dropboxOct 23, 20253 min readCurated summary

With Mobius Labs' Aana models, we're bringing deeper multimodal understanding to Dropbox Dash

Dropbox is integrating Mobius Labs’ Aana multimodal models into Dropbox Dash to improve how it understands text, images, audio, and video together. Aana is designed for efficient, large-scale processing, enabling Dash to interpret scenes, actions, speech, and relationships across media rather than relying on isolated metadata. The goal is to make rich media searchable and support future AI agents that can analyze content and act on it. ## Dropbox Dash’s Multimodal Goal - Dash combines AI assistance and search to understand a team’s work, context, and tools. - Its multimodal capabilities are intended to cover formats such as: - Text - Photos - Videos - Audio - This is especially valuable for creative, technical, media, and marketing teams whose work is distributed across many applications and content repositories. ## Why Multimedia Understanding Is Difficult - A video may require analysis of: - Scene changes - Speaker shifts - On-screen text - Objects and actions - Audio cues and dialogue - Images require systems to understand people, locations, events, and visual context. - Each modality has different timelines and semantic boundaries; transcripts, shots, visuals, and audio do not automatically align. - Effective understanding requires connecting modalities—for example, linking what someone says to the diagram they are showing. - Processing this information across exabytes of content would be prohibitively expensive without highly efficient infrastructure. ## How Aana Combines Models and Context - Aana analyzes images, audio, video, and related text as a connected system instead of separate data streams. - It combines fine-tuned, open-source foundation models for: - Speech, including optimized Whisper-based models such as `faster-whisper-large-v3-turbo` - Vision - Language - Transformer and mixture-of-experts architectures support fast inference on standard GPUs. - Aana tracks how scenes evolve over time, including: - Object movement - Actions - Changing layouts - Cross-modal events - Insights are stored in a shared vector space, enabling searches such as “the part where the presenter explains the API flow” rather than requiring users to scrub through a video manually. ## Efficiency at Dropbox Scale - Aana uses low-bit 8-bit and 4-bit inference through its HQQ system to reduce memory and compute requirements. - Gemlite provides custom GPU kernels to accelerate matrix multiplication and attention operations. - The Aana SDK coordinates models, batching, GPU utilization, and production pipelines. - Teams can configure and deploy different combinations of models while experimenting with new workflows. - These optimizations are intended to make analysis of exabytes of media possible with a much smaller compute footprint than conventional approaches. ## Future Applications in Dash - Aana could help users find specific visual elements across large creative archives. - It could summarize years of client meetings into searchable highlights. - Its capabilities may support agentic workflows that: - Analyze multimedia - Surface relevant insights automatically - Generate ideas - Take actions for teams Dropbox’s integration of Aana is intended to make Dash a more context-aware search and AI assistant for rich media. The practical outcome is less time spent organizing and manually reviewing content, and more time spent creating and acting on information.

Read original

Multimodal AI Transformer Agentic Workflows Vector Db+4

slackOct 23, 20253 min readCurated summary

Advancing Our Chef Infrastructure: Safety Without Disruption

Slack chose to improve its existing Chef and EC2 infrastructure rather than migrate to Policyfiles, avoiding disruptive cookbook and role changes. The central strategy is to divide production into six Availability Zone–based Chef environments, limiting deployment blast radius while preserving existing workflows. A canary environment and staggered release train provide earlier detection of configuration problems and safer fleet-wide rollouts. ## Why Slack Avoided Policyfiles - Policyfiles could have improved long-term safety by replacing roles and environments. - Migrating dozens of teams and their cookbooks would have required substantial effort. - Slack concluded that the short-term disruption and migration risk outweighed the benefits. - Instead, the team extended its existing EC2 framework without requiring cookbook or role changes. ## Splitting Production Chef Environments - Previously, all production nodes used one shared Chef environment. - Cron-triggered Chef runs were staggered across Availability Zones to prevent simultaneous fleet-wide changes. - This reduced the impact of bad changes on existing nodes, but newly provisioned instances immediately consumed the latest version from the shared environment. - During large scale-out events, a broken configuration could therefore spread rapidly to many new nodes. - Slack split production into six environments: `prod-1` through `prod-6`. - Service teams still launch instances as `prod`; internally, nodes are assigned to a numbered environment based on their Availability Zone. - Updates to one environment now affect only the nodes mapped to that environment. ## Extending Poptart Bootstrap - Slack’s base AMIs include `Poptart Bootstrap`, which runs through `cloud-init` during instance startup. - It creates the node’s Chef object, configures DNS, and posts success or failure notifications to Slack. - Slack extended it to inspect the node’s AZ ID and select the appropriate numbered production environment. - This automatically distributes new nodes across isolated Chef environments without requiring service teams to change their provisioning process. ## Canary Deployments and the Release Train - Cookbook changes are promoted: - To sandbox at the top of the hour - To development environments through a Kubernetes cron job - To production beginning at 30 minutes past the hour - `prod-1` acts as the canary production environment. - It receives the latest changes hourly when new cookbook artifacts exist. - This tests changes in real production conditions soon after they are created. - `prod-2` through `prod-6` follow a release train. - A version advances gradually through the production environments. - The next rollout begins only after the previous version has reached `prod-6`. - This sequencing limits the number of affected nodes and makes regressions easier to identify. ## Why `prod-1` Updates Frequently - If the canary waited until a version had passed through every production environment, it would test artifacts containing larger batches of accumulated changes. - Updating `prod-1` frequently keeps the feedback loop close to the originating change. - The remaining production environments provide progressively broader validation after the canary stage. - For example, a new artifact can move from sandbox and dev to `prod-1`, then advance through `prod-2` to `prod-6` while newer artifacts continue entering the canary path. Slack’s approach preserves its existing Chef ecosystem while adding isolation, automated environment assignment, and staged promotion. The result is a safer deployment process that reduces blast radius and catches production issues earlier without forcing widespread application changes.

Read original

Kubernetes Aws Ec2 Chef Canary Deployment+3

figmaOct 22, 20251 min readCurated summary

Visibility at Scale: How Figma Detects Sensitive Data Exposure | Figma Blog

The provided content includes only the title, “How we built a custom permissions DSL at Figma,” and category labels. It does not contain the blog post’s body, so there is not enough information to accurately summarize its technical arguments, implementation details, or conclusions. Please provide the full article text or a link to the post.

Read original

Access Control Authorization Infrastructure Dsl+2

googleOct 22, 2025Original article

Google Earth AI: Unlocking geospatial insights with foundation models and cross-modal reasoning (opens in new tab)

Google Earth AI introduces a framework of geospatial foundation models and reasoning agents designed to solve complex, planetary-scale challenges through cross-modal reasoning. By integrating Gemini-powered orchestrators with specialized imagery, population, and environmental models, the system deconstructs multifaceted queries into actionable multi-step plans. This approach enables a holistic understanding of real-world events, such as disaster response and disease forecasting, by grounding AI insights in diverse, grounded geospatial data. ## Geospatial Reasoning Agents * Utilizes Gemini models as intelligent orchestrators to manage complex queries that require data from multiple domains. * The agent deconstructs a high-level question—such as predicting hurricane landfalls and community vulnerability—into a sequence of smaller, executable tasks. * It executes these plans by autonomously calling specialized foundation models, querying vast datastores, and utilizing geospatial tools to fuse disparate data points into a single, cohesive answer. ## Remote Sensing and Imagery Foundations * Employs vision-language models and open-vocabulary object detection trained on a large corpus of high-resolution overhead imagery paired with text descriptions. * Enables "zero-shot" capabilities, allowing users to find specific objects like "flooded roads" or "building damage" using natural language without needing to retrain the model for specific classes. * Technical evaluations show a 16% average improvement on text-based image search tasks and more than double the baseline accuracy for detecting novel objects in a zero-shot setting. ## Population Dynamics and Mobility * Focuses on the interplay between people and places using globally-consistent embeddings across 17 countries. * Includes monthly updated embeddings that capture shifting human activity patterns, which are essential for time-sensitive forecasting. * Research conducted with the University of Oxford showed that incorporating these population embeddings into a Dengue fever forecasting model in Brazil improved the R² metric from 0.456 to 0.656 for long-range 12-month predictions. ## Environmental and Disaster Forecasting * Integrates established Google research into weather nowcasting, flood forecasting, and wildfire boundary mapping. * Provides the reasoning agent with the data necessary to evaluate environmental risks alongside population density and infrastructure imagery. * Aims to provide Search and Maps users with real-time, accurate alerts regarding natural disasters grounded in planetary-scale environmental data. Developers and enterprises looking to solve high-level geospatial problems can now express interest in accessing these capabilities through Google Earth and Google Cloud. By leveraging these foundation models, organizations can automate the analysis of satellite imagery and human mobility data to better prepare for environmental and social challenges.

AI Gemini Foundation Models Embeddings+5

figmaOct 22, 20252 min readCurated summary

Three New Features to Help Build Brand Momentum With Figma Buzz | Figma Blog

Figma is expanding Figma Buzz to help marketing teams create branded assets at scale while preserving design-system consistency. The October 2025 update introduces configurable templates based on component properties, plugin integrations, and built-in video trimming. Together, these features give marketers more flexibility while helping brand teams maintain control over approved designs, assets, and content. ## Configurable Templates Built on Design Systems - Brand designers can copy templates from Figma Design component sets into Figma Buzz. - Component properties automatically remain functional in Buzz. - Marketers can customize templates through the **Change layout** panel by: - Selecting layout variants - Toggling elements such as dates on or off - Switching between approved component instances - Changing colors, logos, calls to action, or illustrations - Teams can offer flexible template variations without creating and maintaining hundreds of separate files. - The result is easier template discovery and customization without compromising brand guidelines. ## Plugins for Connected Marketing Workflows Plugins are now available directly in Figma Buzz, reducing the need to move between design and marketing tools. Initial integrations focus on digital asset management, translation management, and animation, while teams can also create and publish internal plugins. ### Digital Asset Management DAM integrations let marketers access approved assets and keep designs synchronized without leaving Figma Buzz. Featured plugins include: - Air - Brandfolder - Bynder - OrangeLogic These connections support browsing and importing approved images, as well as uploading completed designs back to asset-management systems. ### Translation Management Translation-management integrations allow teams to pull localized copy directly into Figma Buzz templates. This helps marketing organizations scale campaigns globally while keeping translations connected to their visual design context. ## Video Trimming - Users can trim videos directly inside Figma Buzz. - This provides more control over how video content appears in branded assets. - It builds on earlier support for importing and exporting video. Figma positions these additions as part of its broader effort to make Buzz a scalable workspace for brand and marketing teams. Teams can use design-system-driven templates for consistency, plugins to connect existing workflows, and in-app video editing to produce polished campaign assets more efficiently.

Read original

Figma Design Systems Plugins Figma Buzz+4

dropboxOct 22, 20253 min readCurated summary

Half-Quadratic Quantization of large machine learning models

Half-Quadratic Quantization (HQQ) is a calibration-free method for compressing large machine learning models while retaining quality comparable to calibration-based techniques such as GPTQ and AWQ. It minimizes weight reconstruction error rather than activation error and uses a sparsity-promoting $l_p$ loss to better handle outliers. Because HQQ relies on closed-form alternating updates instead of gradient-based optimization, it can quantize models dramatically faster—reportedly processing Llama-2-70B in under five minutes. ## Why Quantization Matters - Large language models require substantial memory for training and inference. - Methods such as bitsandbytes, GPTQ, and AWQ make models like Llama-2 usable on consumer GPUs. - Weight-only quantization approaches fall into two groups: - **Calibration-free methods**, such as bitsandbytes, use only model weights. - **Calibration-based methods**, such as GPTQ and AWQ, use external datasets. - Calibration-based approaches can provide better quality but: - Their results may depend on calibration-data bias. - Calibration can be computationally expensive for very large models. ## HQQ’s Quantization Objective - Standard quantization can significantly distort weights, particularly outliers with unusually large values. - GPTQ and AWQ reduce the effect of these distortions by minimizing layer-output or activation error using calibration data. - HQQ instead minimizes reconstruction error directly in the weights. - It uses a sparsity-promoting $l_p$ loss, especially with $p<1$, to model heavy-tailed outlier errors more effectively than squared error. - Quantization is defined using: - A scale $s$ - A zero-point $z$ - A quantization operator $Q_{z,s}(W)=\text{round}(W/s+z)$ - A dequantization operator $Q^{-1}_{z,s}(W_q)=s(W_q-z)$ - HQQ fixes the scale and optimizes the zero-point, simplifying the optimization problem. ## Half-Quadratic Optimization - Since the $l_p$ objective with $p<1$ is non-convex, HQQ introduces an auxiliary error variable $W_e$. - The resulting problem is solved through alternating optimization: - Update $W_e$ while holding $z$ fixed. - Update $z$ while holding $W_e$ fixed. - Increase a positive penalty parameter $\beta$ by a factor $\kappa$ each iteration. - This decomposition turns the original difficult problem into simpler sub-problems with closed-form solutions. ## Solving the Sub-Problems - The $W_e$ update is a proximal operation. - For $l_1$ regularization, it corresponds to soft thresholding. - HQQ uses a generalized soft-thresholding operator for $0\leq p\leq1$: \[ \text{shrink}_{l_p}(x,\beta) =\text{sign}(x)\,\text{relu}\left(|x|-\frac{|x|^{p-1}}{\beta}\right) \] - The zero-point update: - Recomputes quantized weights using the current zero-point. - Calculates the difference between quantized weights and corrected original weights. - Sets the new zero-point to the average over the quantization grouping axis. - The implementation optimizes the inverse scale $1/s$, which is more numerically stable in half-precision arithmetic. ## Speed and Practical Advantages - HQQ uses closed-form updates rather than gradients or automatic differentiation. - Quantization can run in inference mode with half-precision arithmetic. - The solver typically converges in only a few iterations. - In contrast, AdamW with PyTorch autograd may require thousands of iterations and fails when using $p<1$. - The article reports HQQ as: - More than 100 times faster than autograd for quantizing Llama-2-7B. - More than 50 times faster than GPTQ for Llama-2-70B. - Capable of quantizing the largest models in only a few minutes. - A 2-bit HQQ version of Llama-2-70B reportedly outperforms full-precision Llama-2-13B at a comparable memory footprint. HQQ is therefore presented as a practical alternative to calibration-based quantization: it combines calibration-free operation and very high speed with competitive compression quality, making rapid experimentation and deployment of large models more feasible.

Read original

Large Language Models Machine Learning Inference Optimization Model Quantization+4

googleOct 21, 2025Original article

A verifiable quantum advantage (opens in new tab)

Google Quantum AI researchers have introduced "Quantum Echoes," a new algorithm designed to measure Out-of-Time-Order Correlators (OTOCs) to characterize quantum chaos. By demonstrating this task on the 103-qubit Willow chip, the team has achieved a verifiable quantum advantage that surpasses the limitations of previous random circuit sampling techniques. This work establishes a direct path toward solving practical problems in physics and chemistry, such as Hamiltonian learning, through the use of stable and reproducible quantum expectation values. ## Limitations of Random Circuit Sampling * While the 2019 "quantum supremacy" milestone proved quantum computers could outperform classical ones, the bitstring sampling method used was difficult to verify and lacked practical utility. * In large-scale quantum systems, specific bitstrings rarely repeat, which restricts the ability to extract useful, actionable information from the computation. * The Quantum Echoes approach shifts focus to quantum expectation values—such as magnetization, density, and velocity—which remain consistent across different quantum computers and are computationally verifiable. ## The Quantum Echoes Algorithm and OTOCs * The algorithm measures OTOCs, which represent the state of a single qubit after a series of "forward" ($U$) and "backward" ($U^\dagger$) evolutions. * In the experiment, 103 qubits on the Willow processor underwent evolution through random quantum circuits to reach a highly chaotic state. * A perturbation (gate $B$) is applied between the forward and backward evolutions; if the system is chaotic, this small change triggers a "butterfly effect," resulting in a final state significantly different from the initial one. * Higher-order OTOCs involve multiple "round trips" of these evolutions, increasing the system's sensitivity to the perturbation and allowing for a more detailed characterization of the quantum dynamics. ## Many-Body Interference and Signal Amplification * The researchers discovered that higher-order OTOCs function like many-body interferometers, where the quantum states of many particles interfere with one another. * The perturbation gates ($B$ and $M$) act as mirrors; when a resonance condition is met (where $U^\dagger$ is the exact inverse of $U$), constructive interference occurs. * This constructive interference amplifies specific quantum correlations, allowing the OTOC signal magnitude to scale as a negative power of the system size, rather than the exponential decay typically seen in chaotic systems. * This amplification makes the OTOC a sensitive instrument for identifying the specific correlations generated between two different qubits during the evolution of the circuit. ## Practical Applications and Future Research The success of the Quantum Echoes algorithm on the Willow chip marks a transition toward using quantum computers for tasks that are both beyond-classical and physically relevant. This method is particularly well-suited for Hamiltonian learning in Nuclear Magnetic Resonance (NMR) and studying the flow of electrons in high-temperature superconductors. Moving forward, the ability to measure verifiable expectation values in the chaotic regime will be essential for researchers looking to simulate complex quantum materials that are impossible to model on classical hardware.

Quantum Computing Quantum Algorithms Quantum Advantage Out Of Time Order Correlators+3

netflixOct 21, 2025Original article

Behind the Streams: Real-Time Recommendations for Live Events Part 3 | by Netflix Technology Blog | Netflix TechBlog (opens in new tab)

Netflix manages the massive surge of concurrent users during live events by utilizing a hybrid strategy of prefetching and real-time broadcasting to deliver synchronized recommendations. By decoupling data delivery from the live trigger, the system avoids the "thundering herd" effect that would otherwise overwhelm cloud infrastructure during record-breaking broadcasts. This architecture ensures that millions of global devices receive timely updates and visual cues without requiring linear, inefficient scaling of compute resources. ### The Constraint Optimization Problem To maintain a seamless experience, Netflix engineers balance three primary technical constraints: time to update, request throughput, and compute cardinality. * **Time:** The specific duration required to coordinate and push a recommendation update to the entire global fleet. * **Throughput:** The maximum capacity of cloud services to handle incoming requests without service degradation. * **Cardinality:** The variety and complexity of unique requests necessary to serve personalized updates to different user segments. ### Two-Phase Recommendation Delivery The system splits the delivery process into two distinct stages to smooth out traffic spikes and ensure high availability. * **Prefetching Phase:** While members browse the app normally before an event, the system downloads materialized recommendations, metadata, and artwork into the device's local cache. * **Broadcasting Phase:** When the event begins, a low-cardinality "at least once" message is broadcast to all connected devices, triggering them to display the already-cached content instantaneously. * **Traffic Smoothing:** This approach eliminates the need for massive, real-time data fetches at the moment of kickoff, distributing the heavy lifting of data transfer over a longer period. ### Live State Management and UI Synchronization A dedicated Live State Management (LSM) system tracks event schedules in real time to ensure the user interface stays perfectly in sync with the production. * **Dynamic Adjustments:** If a live event is delayed or ends early, the LSM adjusts the broadcast triggers to preserve accuracy and prevent "spoilers" or dead links. * **Visual Cues:** The UI utilizes "Live" badging and dynamic artwork transitions to signal urgency and guide users toward the stream. * **Frictionless Playback:** For members already on a title’s detail page, the system can trigger an automatic transition into the live player the moment the broadcast begins, reducing navigation latency. To support global-scale live events, technical teams should prioritize edge-heavy strategies that pre-position assets on client devices. By shifting from a reactive request-response model to a proactive prefetch-and-trigger model, platforms can maintain high performance and reliability even during the most significant traffic peaks.

Distributed Systems Recommendation Systems Caching Scalability+4

datadogOct 21, 20252 min readCurated summary

Detecting malicious pull requests at scale with LLMs | Datadog

Malicious pull requests can turn routine code review and CI workflows into supply-chain attack vectors. The post explains how attackers abuse automated builds—especially when workflows expose repository secrets or elevated GitHub permissions—and recommends treating all pull-request code as untrusted. Strong isolation, least privilege, careful workflow design, and monitoring are essential to prevent credential theft and unauthorized access. ## How Malicious Pull Requests Work - Attackers submit seemingly harmless changes that alter: - GitHub Actions workflows - Build or test scripts - Dependency configuration - Developer tooling - The malicious code executes automatically when CI runs the pull request. - Its goal may be to: - Exfiltrate repository or cloud credentials - Modify artifacts - Access internal systems - Establish persistence in the development pipeline ## Why CI Workflows Are Vulnerable - Pull-request jobs often execute attacker-controlled code through tests, package installation, or build commands. - Using privileged workflow events such as `pull_request_target` can expose secrets while checking out untrusted contributor code. - Broad `GITHUB_TOKEN` permissions increase the impact of a compromised job. - Secrets may leak through logs, environment variables, artifacts, or outbound network requests. ## Defensive Engineering Practices - Treat code from forks and external contributors as untrusted. - Avoid making secrets available to pull-request jobs. - Use minimal `GITHUB_TOKEN` permissions and separate privileged workflows from validation workflows. - Pin third-party GitHub Actions and dependencies to trusted commits or versions. - Require explicit approval before running workflows from untrusted contributors. - Isolate CI jobs with ephemeral runners, restricted network access, and limited filesystem permissions. - Review changes to workflow files with heightened scrutiny. ## Detection and Response - Monitor workflow behavior for unexpected network connections, credential access, or modified build outputs. - Audit repository and CI permissions regularly. - Use short-lived credentials and OIDC-based cloud access instead of long-lived static secrets. - Preserve workflow logs and artifacts to support investigation. - Revoke credentials immediately if a pull request or CI job is suspected of compromise. The practical recommendation is to design CI as though every pull request could be hostile: validate untrusted code in a restricted environment, keep secrets and write permissions out of those jobs, and require deliberate promotion into trusted workflows.

Read original

Kubernetes Datadog Observability Application Performance Monitoring+4

datadogOct 21, 20253 min readCurated summary

Detecting malicious pull requests at scale with LLMs

Datadog built an LLM-powered security reviewer, BewAIre, to examine every pull request for malicious intent as AI-assisted development increases code volume and reviewer fatigue. Unlike traditional static analysis, the system evaluates the design and likely purpose of changes, not just known syntactic patterns. In testing, it achieved over 99.3% accuracy with a 0.03% false-positive rate and now operates in production. ## Why Conventional Review Is Struggling - Datadog processes nearly 10,000 pull requests per week across internal and external repositories, with volume growing as coding assistants become commonplace. - More changes increase the attack surface and make subtle attacks harder to notice. - Reviewers, linters, and scanners can be overwhelmed by verbose or repetitive AI-generated code. - Static analysis is effective at identifying known bad patterns but generally lacks understanding of developer intent. - Legitimate security-sensitive changes, such as permission-model updates, can resemble malicious behavior. ## The tj-actions/changed-files Attack - An attacker used a compromised personal access token belonging to `tj-actions-bot` to modify `index.js`. - The injected Python code extracted secrets and other sensitive information from a GitHub Actions runner’s process memory. - Base64 encoding concealed both the malicious code reference and output written to build logs. - The commit was made to look like a legitimate `renovate[bot]` update. - Version tags were redirected to the malicious commit, allowing the compromise to spread quickly. - Even protected branches, two-person reviews, and other standard controls could miss an attack disguised as a routine dependency or maintenance change. ## BewAIre’s LLM-Based Review Pipeline - **Ingestion:** Collects pull requests merged into the default branch across Datadog repositories. - **Preprocessing:** Normalizes each pull request, extracts its diff, and adds metadata such as the author and repository type. - **Inference:** Sends the code changes and contextual information to an LLM to reason about intent and classify the change as benign or malicious. - **Security signals:** Converts each verdict into a Datadog security signal that appears in internal dashboards and can trigger alerts or pages for security engineers. - Each finding includes a written explanation of why the change may be malicious. ## Improving Accuracy and Limiting Noise - Datadog used prompt engineering and data tuning to improve the model’s ability to distinguish malicious behavior from legitimate changes. - Suppression rules were added for known-safe patterns. - Testing on a curated, balanced dataset of hundreds of malicious and benign pull requests produced: - More than 99.3% overall accuracy. - More than 93.5% balanced accuracy. - A 0.03% false-positive rate. - The team also designed the system to address model degradation caused by context limits and continuously evaluated it against real-world attacks, including the tj-actions and Nx incidents. ## Production Deployment - BewAIre reviews Datadog pull requests in real time and is already deployed across the company’s repositories. - The system is available in Preview to Static Code Analysis customers. - Its purpose is to add scalable, intent-focused detection without imposing stricter manual review requirements that could slow development. Datadog’s approach suggests that LLMs can complement—not replace—static analysis and human investigation by providing continuous, intent-aware security review at large scale.

Read original

Large Language Models Prompt Engineering Code Review Git+4

figmaOct 20, 20252 min readCurated summary

Design Systems: From the Basics to Big Things Ahead | Figma Blog

Design systems create consistency at scale while connecting design work to production. The post argues that as AI accelerates product development, organizations need a well-documented, shared foundation more than ever. It offers a progression from design-system fundamentals to adoption and measurement, alongside examples and resources for teams at different maturity levels. ## Design Systems 101 - Design systems evolved from: - Graphic-design and print-era style guides - Typographic standards and brand guidelines - Digital systems created by companies such as IBM, Microsoft, Apple, and Google - These systems translated visual and interaction standards from paper into digital interfaces. - The National Park Service’s adaptation of Massimo Vignelli’s 1977 design system illustrates how established systems can be carried into modern digital products. - Figma’s design-system series is intended for both teams starting their first system and organizations scaling existing ones. ## Documentation Drives Adoption - Documentation turns abstract principles into practical guidance for designers and developers. - It provides a shared reference point and helps users understand how to work with system tools and components. - Alaska Airlines prioritized documentation for its Auro design system, including guidance for Figma features such as auto layout and branch merging. - Documentation must serve different roles: - Some users need detailed specifications. - Others benefit from high-level explanations or visual examples. - As more roles participate in product design, documentation needs to remain clear, accessible, and relevant. ## Measuring Design-System ROI - Organizations can evaluate business impact through: - Component usage - Adoption rates - Consistency scores - Metrics can reveal not only whether a system is being used, but also where it needs improvement. - At athenahealth, increased detachments from a container component prompted investigation. - Detaching may indicate: - A component bug - Missing functionality - Unexpected combinations of existing elements - These signals help design-system teams prioritize fixes and enhancements. The practical recommendation is to treat a design system as an evolving product: establish strong foundations, document them for varied audiences, and use adoption and usage data to continually improve the system.

Read original

Figma AI Design Systems Design Tokens+3

googleOct 19, 2025Original article

A picture's worth a thousand (private) words: Hierarchical generation of coherent synthetic photo albums (opens in new tab)

Researchers at Google have developed a hierarchical method for generating differentially private (DP) synthetic photo albums, providing a way to share representative datasets while protecting sensitive individual information. By utilizing an intermediate text representation and a two-stage generation process, the approach maintains thematic coherence across multiple images in an album—a significant challenge for traditional synthetic data methods. This framework allows organizations to apply standard, non-private analytical techniques to safe synthetic substitutes rather than modifying every individual analysis method for differential privacy. ## The Hierarchical Generation Process * The workflow begins by converting original photo albums into structured text; an AI model generates detailed captions for each image and a summary for the entire album. * Two large language models (LLMs) are privately fine-tuned using DP-SGD: the first is trained to produce album summaries, and the second generates individual photo captions based on those summaries. * Synthetic data is then produced hierarchically, where the model first generates a global album summary to serve as context, followed by a series of individual photo captions that remain consistent with that context. * The final step uses a text-to-image AI model to transform the private, synthetic text captions back into a set of coherent images. ## Benefits of Intermediate Text Representations * Text summarization is inherently privacy-enhancing because it is a "lossy" operation, meaning the text description is unlikely to capture the exact unique details of an original photo. * Using text as a midpoint allows for more efficient resource management, as generated albums can be filtered and curated at the text level before undergoing the computationally expensive process of image generation. * The hierarchical approach ensures that photos within a synthetic album share the same characters and themes, as every caption in a set is derived from the same contextual summary. * Training two separate models with shorter context windows is significantly more efficient than training one large model, because the computational cost of self-attention scales quadratically with the length of the context. This hierarchical, text-mediated approach demonstrates that high-level semantic information and thematic coherence can be preserved in synthetic datasets without sacrificing individual privacy. Organizations should consider this workflow—translating complex multi-modal data into structured text before synthesis—to scale differentially private data generation for advanced modeling and analysis.

AI Large Language Models Generative AI Differential Privacy+4