techlist.io
KREN

Database Design

191 posts

aws4 min readCurated summary

20 years in the AWS Cloud – how time flies! | Amazon Web Services

AWS’s 20-year evolution reflects a shift from foundational cloud infrastructure to managed services for AI, automation, and agentic applications. The author argues that AWS’s most important innovations come from responding to customer needs rather than chasing every fashionable technology. Personal experiences with AWS and its community illustrate how cloud services have enabled developers, researchers, and businesses to pursue previously impractical projects. ## AWS’s Impact on the Author’s Career - The author met AWS blogger Jeff Barr in Seoul in 2006, shortly after Amazon began promoting API-based services. - Inspired by Barr, the author began building APIs for third-party developers and later used AWS for large-scale academic research. - The author’s company became one of Korea’s earliest AWS customers in 2014. - AWS helped make advanced computing capabilities accessible to individuals, startups, researchers, and enterprises. ## Innovation Driven by Customer Needs - AWS has grown to more than 240 cloud services and launches thousands of features each year. - The author highlights the importance of distinguishing genuine technological trends from temporary distractions. - AWS’s evolution spans deep learning, generative AI based on large language models, and today’s agentic AI. - The central innovation principle is to listen to customers and solve their most important problems, rather than adopting technology simply because it is fashionable. ## Major AWS Milestones The article recalls foundational services from AWS’s first decade, including: - Amazon S3 and EC2 in 2006 - Amazon RDS and VPC in 2009 - DynamoDB and Redshift in 2012 - WorkSpaces and Kinesis in 2013 - AWS Lambda in 2014 - AWS IoT in 2015 ## Containers and Serverless Databases - Amazon ECS, launched in 2014, simplified running containers across managed EC2 clusters. - Amazon EKS later added managed Kubernetes, while AWS Fargate enabled serverless container deployment. - Amazon Aurora provided highly available relational databases at scale. - Aurora Serverless evolved from version 1 to version 2, which can scale down to zero. - Aurora DSQL, launched in 2025, extends the serverless model to distributed SQL workloads requiring continuous availability. ## Making Machine Learning More Accessible - Amazon SageMaker, launched in 2017, provided an end-to-end managed environment for building, training, and deploying ML models. - In 2024, AWS introduced the next-generation SageMaker platform for data, analytics, and AI, along with SageMaker AI for model development and deployment. - AWS also developed specialized hardware: - Inferentia for low-latency inference - Trainium for high-performance AI training - Trainium3 UltraServers for improved economics in generative AI workloads ## Improving Cloud Price Performance - EC2 A1 instances introduced AWS Graviton processors based on Arm architecture. - Later Graviton generations expanded price-performance benefits across services such as ECS, EKS, Lambda, RDS, ElastiCache, EMR, and OpenSearch Service. - More than 90,000 customers have reportedly adopted Graviton-based infrastructure. ## Hybrid Cloud and Edge Computing - AWS Outposts brings AWS infrastructure and services into customer data centers and edge locations. - Available configurations range from 1U and 2U servers to 42U racks and multi-rack deployments. - Customers use Outposts for low-latency access, local processing, data residency, and applications with on-premises dependencies. ## Generative AI and Agentic Development - Amazon Bedrock provides access to multiple AI models and managed capabilities for building secure generative AI applications. - Bedrock AgentCore extends the platform to deploying and operating agents at scale. - More than 100,000 customers use Bedrock for personalization, workflow automation, and insight generation. - Amazon CodeWhisperer evolved into Amazon Q Developer, adding conversational assistance, project-based generation, and code transformation. - The service later evolved into Kiro, an agentic development tool centered on spec-driven development and autonomous coding tasks. - AWS expanded model choice through Amazon Titan and Amazon Nova, including services for building frontier models and browser-automation agents. AWS’s history suggests that the strongest path forward is to use AI and cloud services to address concrete customer and business challenges. The author’s examples present AWS as an evolving platform whose value comes not only from individual launches, but from steadily making advanced infrastructure, machine learning, and autonomous software development more accessible.

Read original(opens in new tab)
Database DesignMachine LearningKubernetesAWS+4
stripe3 min readCurated summary

Testing the impact of Adaptive Pricing across 1.5M subscription checkout sessions

Subscription businesses face major operational and payment challenges when pricing internationally, especially because exchange rates fluctuate across recurring billing cycles. Stripe’s Adaptive Pricing addresses this by displaying and charging customers in local currencies while managing FX conversion and renewal stability. An analysis of 1.5 million checkout sessions found that it increased signup conversion, payment authorization, and subscription lifetime value. ## The Challenge of Localized Subscription Pricing - Localizing prices requires businesses to manage: - FX risk and conversion fees - Currency-specific price lists - Ongoing accounting, reconciliation, and reporting - Subscriptions are more complex than one-time purchases because prices must remain predictable at every renewal. - Cross-border charges are more likely to fail than local-currency transactions. - In 2025, 80% of subscription transactions were still priced in a company’s default currency. ## How Adaptive Pricing Works - Stripe’s Adaptive Pricing, available through the Optimized Checkout Suite, automatically displays prices in a customer’s local currency. - Stripe handles currency conversion and related operational work. - A stability buffer helps keep renewal amounts consistent despite exchange-rate changes. - For example, a Brazilian customer might continue paying R$49.60 per month rather than seeing a different converted amount each billing cycle. - Significant exchange-rate movements can still cause a renewal amount to change, similar to adjustments made by card issuers. ## Measured Signup Improvements Stripe analyzed 1.5 million subscription checkout sessions, comparing Adaptive Pricing with a randomized 1% holdback group. - Average conversion increased by 4.7%. - Average payment authorization increased by 1.9%. - Lifetime value per checkout session increased by 5.4%. - Some businesses saw LTV gains above 30%. - Runway reported a 14% increase in LTV per session and 17.7% more LTV per subscription. Local prices reduce the need for customers to mentally convert costs and make recurring commitments feel more transparent. Charging locally can also improve authorization rates because cross-border payments are more likely to be declined. ## Retention and Long-Term Value - Customers who paid in local currency showed consistently higher retention than those charged in a company’s default currency. - Better signup conversion combined with more successful renewals increases the value generated by each checkout session. - Even small improvements in initial conversion and payment approval can compound into meaningful gains in subscription LTV. ## Scaling Globally - Businesses that do not localize subscription prices may be losing international revenue. - Adaptive Pricing lets companies offer local-currency pricing without building their own FX systems, currency-specific price catalogs, or renewal logic. - Stripe reports that more than 500,000 businesses, including over 16,000 subscription companies such as Cursor, Perplexity, and Runway, use Adaptive Pricing. For subscription companies expanding internationally, localized pricing can improve both initial purchase performance and long-term subscriber value while reducing the complexity of managing global billing.

Read original(opens in new tab)
Database DesignStripeLocalizationAdaptive Pricing+4
line4 min readCurated summary

Utilizing SLI/SLO to Improve Reliability Part 1: SLI/SLO Framework and the Development Story of Service Status Check Tool LINE Status

Repeated SLI/SLO adoption revealed a common process that could be standardized across services. The team turned that process into a reusable framework and built “LINE Status,” an internal tool that automatically presents service health according to user experience rather than raw alerts. Together, these initiatives create a shared organizational language for understanding reliability and its impact on users. ## A Reusable SLI/SLO Framework After applying SLI/SLOs to several platforms and services, the SRE team identified recurring patterns independent of service type. They organized these patterns into a five-stage framework: - **Select critical user journeys (CUJs) and define SLIs** - Identify the experiences most important to users. - Define measurable SLIs that represent those experiences. - **Design instrumentation and metrics** - Build or adapt metrics suitable for each CUJ. - Use standardized naming based on Prometheus or OpenTelemetry. - **Create dashboards and recording rules** - Provide Grafana dashboards for quickly assessing SLO achievement. - Precompute complex PromQL operations to improve query performance. - **Set SLOs and alerts** - Begin with flexible targets, such as 99.9% availability over a 28-day rolling window, allowing roughly 40 minutes of downtime. - Define runbooks for responding to alerts. - Refine targets after operational data and experience accumulate. - **Establish error-budget governance** - Balance release speed against reliability. - Review objectives monthly or quarterly. - Adjust SLOs and processes as needed. The framework is currently distributed as a Confluence template containing guidance and FAQs, reducing the communication effort required from SREs during initial adoption. ## Moving from Alerts to User-Centered Service Status As more services adopted SLI/SLOs, the team wanted a consistent way to understand the health of services they did not directly operate. - The existing public LINE Status API page focused on external users and was updated manually during major incidents. - The new internal tool was intended to: - Represent the status of individual service components. - Update automatically from SLI/SLO alerts and outage data. - Show whether user experience was being affected. - Rather than simply reflecting whether an alert or outage existed, status was based on CUJ-related SLI performance and SLO achievement. - Only representative, high-value CUJs were exposed, avoiding unnecessary technical detail. ## LINE Status Architecture and Interface LINE Status was designed as more than an alert list. It collects events through webhooks, stores them in a separate database, and uses that data to track both current status and historical changes. - Technical SLI/SLO terms are translated into user-facing functions such as “Message Sending” or “Read Receipts.” - Status colors provide an immediate overview: - Green: normal - Yellow: event detected - Red: outage - The main page provides: - An overview of all services. - CUJ status within each service card. - AI-generated one-line summaries. - Service detail pages provide: - Recently affected items near the top. - Timeline-based event displays. - Monthly historical events. - The history page shows: - The scope of impact for each service during an event. - Past events organized by month. The initial implementation took about a month and was refined through colleague feedback. The author also used AI-assisted “vibe coding” for the frontend, emphasizing that clear, detailed requirements were more important than the development tool itself. ## Connecting the Framework and LINE Status Once a service adopts SLI/SLOs through the framework, it can be registered in LINE Status. This connects the definition of reliability objectives with an organization-wide view of service health. - Developers and operators can use the same CUJ-based standards. - Teams can focus on whether users are affected instead of interpreting isolated alerts. - During incidents, the tool helps identify impacted experiences quickly. - Over time, the approach may improve decision-making speed and cross-team communication. The team plans to refine CUJs, SLIs, and status-transition rules through continued operational experience. The practical goal is to make SLI/SLOs a common language for describing service health, enabling reliability practices to scale without depending heavily on individual teams or specialists.

Read original(opens in new tab)
Database DesignAPIPrometheusOpentelemetry+3
gitlab3 min readCurated summary

GitLab 18.10: Agentic AI now open to even more teams on GitLab

GitLab 18.10 makes agentic AI available to Free GitLab.com teams without requiring a subscription upgrade. By purchasing shared monthly GitLab Credits, teams gain access to planning, code generation, automated code review, and pipeline troubleshooting. The update also introduces predictable flat-rate pricing for code reviews, while Premium remains attractive for teams needing broader platform capabilities and included credits. ## Agentic AI for Free-tier teams - Free top-level GitLab.com groups can purchase a monthly commitment of GitLab Credits through group billing. - Credits are shared across the entire team, so organizations pay for AI usage rather than per-user access. - Teams receive access to capabilities previously available to Premium and Ultimate customers, including: - Planner Agent - Developer Flow - Code Review Flow - Fix CI/CD Pipeline Flow - Agentic Chat - Code Suggestions - Custom agents and flows - Group owners can use the GitLab Credits dashboard to monitor which agents and workflows consume credits. ## From planning to deployment GitLab describes a workflow covering the full software lifecycle: - **Planner Agent** turns a natural-language feature request into structured issues with descriptions, labels, and relationships. - **Developer Flow** reads an issue, generates code, runs tests, and opens a merge request. - **Code Review Flow** performs multi-step automated reviews and posts inline feedback based on repository context and code changes. - **Fix CI/CD Pipeline Flow** analyzes failed job logs, identifies likely root causes, and proposes fixes. - Agentic Chat supports iterative tasks such as refactoring, extending, or explaining code. ## Flat-rate automated code review - Code Review Flow costs **0.25 GitLab Credits per review**, regardless of merge request size, repository complexity, or internal processing steps. - Four reviews consume one credit. - The fixed price makes costs easier to forecast for both small and high-volume teams. - Automated reviews can run concurrently, reducing review queues and freeing human reviewers to focus on architecture and business logic. ## Why Premium may be the next step - GitLab Premium costs **$29 per user per month** and includes 12 promotional credits per user. - A 20-person team would receive 240 credits monthly—enough for approximately 960 automated code reviews or a mixture of AI workflows. - Premium also adds advanced CI/CD, merge approvals, code owners, governance features, and unified project context. - Teams that begin with Free plus purchased credits may find Premium more economical as AI becomes central to their development process. ## Getting started Free GitLab.com teams can purchase credits through group billing and begin using the Duo Agent Platform immediately. Teams seeking broader collaboration, governance, and CI/CD features can instead evaluate GitLab Premium or Ultimate.

Read original(opens in new tab)
Database DesignAI AgentsCI/CDGitlab+3
meta3 min readCurated summary

Friend Bubbles: Enhancing Social Discovery on Facebook Reels

Friend bubbles in Facebook Reels surface videos that friends have liked or interacted with, combining content discovery with opportunities for conversation. The system uses machine-learning models to estimate viewer-friend closeness, retrieve relevant friend-interacted videos, and rank them alongside conventional recommendation signals. Its goal is not to show the most bubbles possible, but to identify meaningful connections and content that can drive both engagement and social interaction. ## System Architecture - The recommendation system combines: - **Viewer-friend closeness**, determining whose interactions matter most. - **Video relevance**, determining which friend-interacted videos best fit the viewer. - Multiple friends interacting with the same video can indicate stronger shared interest. - Social discovery and engagement reinforce one another: relevant friend content encourages interaction, which improves the system’s understanding of the social graph. ## Modeling Viewer-Friend Closeness - Facebook uses two complementary models: - A survey-based model estimating real-world relationship strength. - An activity-based model estimating closeness from on-platform behavior. - The survey model considers: - Mutual friends and interaction patterns. - User-provided attributes such as location. - Number of friends and posts shared. - Communication frequency and other survey proxies for offline closeness. - Users are periodically asked whether they feel close to a randomly selected connection. - The model is refreshed regularly and performs weekly inference across trillions of friend relationships. - The activity-based model learns from likes, comments, reshares, and interactions occurring after bubbles are shown. - Facebook prioritizes connection quality over quantity: larger friend networks may create more opportunities, but the system aims to surface only relationships likely to make recommendations meaningful. ## Retrieving and Ranking Friend Content ### Expanding Candidate Retrieval - The retrieval stage explicitly sources videos interacted with by close friends. - This expands the recommendation funnel so high-quality friend content can reach downstream ranking systems. - Without dedicated retrieval, relevant friend videos might never become candidates. ### Adding Social Context to Ranking Models - Friend-interacted videos could rank poorly when models lacked viewer-friend closeness information. - The system added bubble interaction signals and relationship-strength features to early- and late-stage multi-task, multi-label ranking models. - These features help models distinguish social relevance from ordinary content-interest signals. - Feedback from bubble impressions and resulting interactions continuously flows back into model training. - Ranking objectives consider: - Watch time. - Likes and comments. - The probability of engagement after a bubble impression: `P(video engagement | bubble impression)`. - Tunable weights balance entertainment and video quality against social goals such as discovering friends’ interests and encouraging conversation. ## Client Infrastructure and Reels Performance - Friend-bubble metadata had to be integrated without harming Reels’ core experience. - The implementation targeted: - Smooth scrolling. - No additional loading latency. - Low CPU usage during metadata retrieval and processing. - Facebook aligned bubble metadata retrieval with the existing video prefetch window, which already loads metadata, thumbnails, and buffered content before playback. - This allows the system to reuse cached results and avoid adding unnecessary work during scrolling. Friend bubbles work best when social relevance and content quality are optimized together. By combining relationship models, friend-aware retrieval and ranking, feedback-driven learning, and performance-conscious client infrastructure, Facebook turns shared video interests into lightweight opportunities for discovery and conversation.

Read original(opens in new tab)
Database DesignMachine LearningRecommendation SystemsFeedback Loop+3
aws2 min readCurated summary

Our First 2026 Heroes Cohort Is Here! | Amazon Web Services

AWS has announced its first 2026 Heroes cohort, recognizing Maurizio, Ray Goh, and Sheyla Leacock for combining technical expertise with community leadership. Their work spans cloud architecture, generative AI, machine learning, and cybersecurity, while emphasizing mentorship, education, and meaningful human connections. Together, they demonstrate how technology leaders can expand access to skills and strengthen communities globally. ## Maurizio – Pignola, Italy - CTO and organizer of the AWS User Group Basilicata. - Has spent more than a decade developing cloud communities and technology ecosystems in areas where they previously did not exist. - Founded an international technology conference in a small mountain village, connecting global experts with local developers. - Covers topics including cloud architecture, DevOps, and web scaling, alongside creative networking opportunities. - Mentors children, university students, and professionals transitioning into cloud careers. - Combines technical leadership with inclusive, cross-generational community building. ## Ray Goh – Singapore - AI and machine learning community leader involved in AWS programs since 2018. - Founded The Gen-C in 2024, offering public library workshops on generative AI, LLM fine-tuning, and AWS AI agents. - Has spoken at major AWS events and contributed to the AWS Machine Learning Blog. - Led DBS Bank’s AWS DeepRacer initiative, which trained more than 3,100 employees. - Trained over 1,300 ASEAN students in LLM techniques in 2025. - Supports skills-based programs teaching AI and machine learning to women, children, and young people. ## Sheyla Leacock – Panama City, Panama - IT security professional, mentor, technical writer, and international speaker. - Leads the AWS User Group in Panama and participates in AWS Community Days and regional meetups. - Has spoken at AWS Summits, AWS re:Invent PeerTalk sessions, and more than 20 international conferences. - Publishes educational content focused on AWS cloud computing and cybersecurity. - Works with universities as a guest lecturer to help develop future technology and security professionals. - Strengthens the cloud and cybersecurity ecosystem through education, knowledge sharing, and community leadership. The new cohort highlights the broader impact of community-driven technology leadership. Readers can visit the AWS Heroes webpage to learn more about the program or connect with a Hero.

Read original(opens in new tab)
Database DesignAIMachine LearningAWS+4
line4 min readCurated summary

Large-scale iOS Settings System Unraveled through AttributedString Structure

LINE’s “Service Configuration” system lets teams deploy features dynamically without waiting for LINE’s two-week app release cycle. As the iOS app grew to roughly 700 configuration keys across 60 modules, its monolithic design created dependency, usability, concurrency, testing, and QA problems. The article argues that the original design was reasonable at small scale but needed to evolve, beginning with lessons from Foundation’s type-safe `AttributedString` design. ## What Service Configuration Provides - Service operators modify values through an administration page. - The server notifies LINE clients, which fetch updated values. - Values are selected based on factors such as: - User region - Device - OS version - The system supports: - Feature flags - Rollbacks - A/B tests - Error-reporting sample rates - UI behavior policies - Configuration is delivered as a string-to-string dictionary, for example: - `"function.media.image_medium": "1280,70"` - `"function.media.message.flow.v2.image": "Y"` ## Problems Caused by the Monolithic Design The original implementation required every key to be declared in one roughly 7,000-line file. Although this was simple initially, growth in teams and modules made the structure increasingly costly. ### Circular Dependencies and Weak Typing - Configuration values were exposed as raw strings because the configuration module could not depend on feature-specific modules. - For example, `"1280,70"` represented image dimensions and JPEG quality, but callers had to parse it into an `ImageTransferQuality` value themselves. - Defining `ImageTransferQuality` in the configuration module avoided repeated parsing but polluted unrelated modules with photo-specific types. - Defining it in the photo module preserved separation of concerns but created an impossible reverse dependency. ### Incomplete and Confusing Abstractions - Developers had to understand server-specific encoding rules and implementation details. - Boolean values were sent as `"Y"` and `"N"`, requiring a custom `decodeBoolIfPresent(forKey:)` method. - The custom decoder’s name resembled Swift’s standard decoding API, making incorrect implementations easy to write and review. - Decoding failures could silently fall back to defaults, making the underlying problem difficult to diagnose. - The same default value often had to be declared three times: - A property-group default - A decoding fallback - A global `defaultConfiguration` entry - These duplicated defaults served subtly different purposes, although the distinctions were generally unnecessary. ### Lack of Thread Safety - Configuration groups were lazily decoded and replaced when new server values arrived. - Multiple services could read configuration values concurrently on different threads. - This caused use-after-free crashes— reportedly hundreds per day—leading to bug tickets and hotfix releases. - As the number of services and concurrent operations increased, this became a systemic issue rather than an occasional edge case. ### No Built-in Debug Overrides - QA frequently needed to temporarily change configuration values. - Because the system had no override mechanism, each feature required custom: - Persistent storage - Debug-menu UI - Value-display text - Implementing this repeatedly required edits across several files and modules. ### Fragmented Test Doubles - Since `LineConfigurationManager` was a singleton, modules created narrow protocols and custom mocks for the settings they used. - This resulted in dozens of duplicated protocols and test doubles. - These had to be updated alongside configuration keys and could fall out of sync. - Differences between mocks and production behavior could allow bugs to escape tests or create false failures. ## Looking to Established Designs The team first distilled the required properties of a replacement: - Type-safe access to a large number of key-value pairs - Independent key definitions by each module - Safe behavior under concurrency They identified Foundation’s `AttributedString` as a useful precedent because it manages many typed attributes while allowing UIKit, AppKit, SwiftUI, and other frameworks to define their own attributes independently. The article presents this as the starting point for redesigning Service Configuration around a more modular and type-safe architecture.

Read original(opens in new tab)
Database DesignAb TestingSwiftType Safety+4
stripe2 min readCurated summary

Introducing the Machine Payments Protocol

AI agents are moving beyond chatbots toward autonomous systems that plan, act, and evaluate results, creating demand for agent-friendly commerce. Stripe and Tempo are launching the Machine Payments Protocol (MPP), an open standard that lets agents make programmatic payments to businesses and services. MPP supports microtransactions, recurring payments, stablecoins, fiat, and existing Stripe payment methods without requiring human intervention. ## The Challenge of Agent Payments - Traditional financial workflows are designed for humans. - Agents often cannot independently: - Create accounts - Navigate pricing and subscription options - Enter payment details - Configure billing - These obstacles limit agents’ ability to purchase services and participate in the internet economy. ## How the Machine Payments Protocol Works - An agent requests a resource from a service, API, MCP server, or other HTTP endpoint. - The service returns a payment request. - The agent authorizes payment. - The requested resource is delivered automatically. - Stripe businesses can integrate MPP through the PaymentIntents API with only a few lines of code. - Payments appear in Stripe’s existing API and Dashboard and settle through the business’s normal balance, currency, and payout schedule. - Standard Stripe capabilities remain available, including tax calculation, fraud protection, reporting, accounting integrations, and refunds. ## New Agentic Business Models MPP is already enabling agents to pay for services such as: - Browserbase: headless browsers billed per session - PostalForm: printing and mailing physical documents - Prospect Butcher Co.: ordering food for pickup or delivery in New York City - Stripe Climate: making programmatic contributions - Parallel Web Systems: paying per API call for web access Payments can use stablecoins, cards, buy now, pay later methods, and Shared Payment Tokens. ## Stripe’s Agent Economy Infrastructure Stripe positions MPP alongside its broader Agentic Commerce Suite, Agentic Commerce Protocol, MCP integrations, and support for x402. Together, these tools are intended to help businesses sell directly to agents and support new automated commerce patterns. Businesses interested in enabling agent payments can review Stripe’s MPP documentation and join the early-access program.

Read original(opens in new tab)
Database DesignStripeAgentic CommerceMachine Payments Protocol+4
dropbox3 min readCurated summary

How we optimized Dash's relevance judge with DSPy

Dropbox Dash needed a relevance judge that could score query–document pairs accurately, cheaply, and reliably at scale. Its original judge used OpenAI’s o3, but the cost made it impractical for large-scale labeling, while its prompt performed poorly when moved to the cheaper gpt-oss-120b model. Dropbox used DSPy’s GEPA optimizer to turn prompt tuning into a measurable feedback loop, improving alignment with human judgments while preserving production-ready output formatting. ## Measuring Agreement with Human Reviewers - The judge rates each query–document pair on a 1–5 relevance scale: - **5** means a perfect match. - **1** means no meaningful connection to the query or user intent. - Human annotators provide both: - A relevance score. - A short explanation for their judgment. - Dropbox evaluates the model with normalized mean squared error (NMSE): - It measures the squared difference between model and human ratings. - Scores are normalized to a 0–100 scale. - **0** represents perfect agreement; higher values indicate worse performance. - Invalid JSON or incorrectly structured responses are treated as fully incorrect because they cannot be consumed reliably by downstream systems. - The optimization objective is therefore twofold: - Minimize disagreement with human ratings. - Ensure consistently parseable, production-ready outputs. ## Moving from o3 to a Lower-Cost Model - The original judge used OpenAI’s o3 because it delivered strong agreement with human ratings. - Running o3 across orders of magnitude more query–document pairs was too expensive. - Dropbox selected **gpt-oss-120b**, an open-weight model offering a better cost-performance balance. - The carefully tuned o3 prompt did not transfer directly: - Relevance quality declined under the NMSE metric. - Manual prompt rewriting would have required extensive iteration and regression testing. ## DSPy and GEPA-Based Prompt Optimization - Dropbox defined the optimization problem using: - A fixed relevance-rating task. - Human-annotated examples. - NMSE as the evaluation metric. - DSPy’s **GEPA optimizer** iteratively improves prompts for a specific target model. - Instead of relying only on an aggregate score, GEPA analyzes individual disagreements and generates structured feedback. - Feedback combines: - The difference and direction between predicted and human ratings. - The human annotator’s explanation. - The model’s reasoning. - DSPy then uses a reflection loop: - Evaluate the current prompt. - Identify recurring failure modes. - Revise the prompt with generalizable rules. - Repeat the process against the human-alignment metric. - This approach can address systematic errors such as: - Overvaluing keyword overlap. - Undervaluing document recency. - Misinterpreting user intent. - The feedback explicitly discourages overfitting to individual examples and preserves core task constraints, including the 1–5 rating range. Dropbox’s experience suggests that relevance judges should be optimized systematically rather than tuned manually. Defining a clear human-alignment metric, including structural validity, allows DSPy to adapt prompts across models while reducing cost and limiting regressions.

Read original(opens in new tab)
Database DesignJsonOpenaiDspy+3
google3 min readCurated summary

Improving breast cancer screening workflows with machine learning

Google Research’s AIMS studies evaluated whether machine learning could support the UK’s mammography double-reading workflow. Across five NHS screening services, the AI system improved cancer detection sensitivity without reducing specificity, detected some cancers missed by human readers, and processed cases far faster. The studies also showed that safe deployment requires local calibration, monitoring for distribution shifts, and evaluation of how clinicians interact with AI results. ## NHS Screening Challenges - The UK NHS uses two human readers for each mammogram, with arbitration when their assessments require review. - A projected shortage of clinical radiologists—currently around 30% and expected to reach 40% by 2028—threatens the sustainability of this model. - AI could help increase detection while reducing pressure on radiology services. ## Study 1: Standalone Performance - The retrospective evaluation included mammograms from approximately 116,000 women screened across five NHS services. - The services represented three different double-reading and arbitration workflows. - AI thresholds were calibrated separately for each service to account for local populations and procedures. - Performance was measured against the original first reader using a 39-month follow-up period, including interval and subsequent-round cancers. - Researchers also assessed: - Comparisons with second and consensus readers - Lesion-level localization - Performance across demographic groups ## Study 1: Results - Cancer detection increased from 7.54 to 9.33 cases per 1,000 women. - The AI system achieved significantly higher sensitivity than the original first reader without compromising specificity. - It detected 25% of interval cancers missed by the original double-reading process. - Performance was especially strong for invasive cancers and women attending their first screening. - The study found no notable systematic disparities by age, ethnicity, breast density, or socioeconomic status. ## Prospective Technical Deployment - The system was deployed non-interventionally at 12 sites across two London screening services. - It processed 9,266 cases over roughly two months per service. - Mammograms were pseudonymized and sent to a secure Google Cloud-based system. - Median AI processing time was 17.7 minutes, compared with more than two days for the first human read. - The deployment detected a distribution shift between historical training data and current clinical data. - Researchers adjusted operating points during deployment to maintain safe and appropriate recall rates for local workflows. ## Study 2: AI in the Double-Reading Workflow - The second study examined how human readers performed when using AI as part of arbitration, rather than evaluating AI in isolation. - Twenty-two readers reviewed thousands of cases using real screening-service rules. - Two workflows were compared: - **Standard care:** decisions from the historical first and second human readers - **AI-enabled care:** the historical first-reader decision paired with the AI decision - This design aimed to assess the practical effects of replacing the second human read with an AI reader. The findings support AI as a potential second reader in breast cancer screening, but broader prospective clinical validation is still needed. Successful adoption should include phased deployment, local calibration, continuous monitoring, and careful evaluation of human-AI decision-making.

Read original(opens in new tab)
Database DesignAIMachine LearningCloud Computing+4
aws3 min readCurated summary

AWS Weekly Roundup: Amazon S3 turns 20, Amazon Route 53 Global Resolver general availability, and more (March 16, 2026) | Amazon Web Services

Amazon S3 marked its 20th anniversary with major milestones in scale, performance, and cost reduction, while AWS introduced account regional namespaces for improved bucket-name control. The week’s featured launch was the general availability of Amazon Route 53 Global Resolver, providing secure, globally accessible DNS resolution across 30 AWS Regions. Other updates covered stateful AI agent infrastructure, Windows Server 2025, simplified AWS identity access, and reusable Redshift ingestion templates. ## Amazon S3 Reaches 20 Years - Launched publicly on March 14, 2006, S3 has grown from object storage into a foundational cloud service. - As of March 2026, it stores: - More than 500 trillion objects - Hundreds of exabytes of data - Over 200 million requests per second globally - Storage prices have fallen by approximately 85% since launch, to just over $0.02 per gigabyte. - New account regional namespaces let organizations reserve bucket names within their own account namespace by adding an account-specific suffix. - Adoption can be enforced with IAM and AWS Organizations service control policies using the `s3:x-amz-bucket-namespace` condition key. ## Route 53 Global Resolver Becomes Generally Available - Amazon Route 53 Global Resolver is an internet-reachable, anycast DNS resolver for authorized clients anywhere. - It is available across 30 AWS Regions and supports IPv4 and IPv6 DNS queries. - It resolves: - Public internet domains - Private domains associated with Route 53 private hosted zones - Security features include filtering for malicious, unsafe, DNS tunneling, and Domain Generation Algorithm (DGA) domains. - General availability adds protection against Dictionary DGA threats. - Centralized DNS query logging is also included. ## Additional AWS Service Updates - **Bedrock AgentCore Runtime** - Adds stateful MCP server support through the `Mcp-Session-Id` header. - Dedicated microVMs isolate each user session and preserve context across interactions. - MCP servers can use elicitation, sampling, and progress notifications in addition to resources, prompts, and tools. - **Amazon WorkSpaces** - Adds Windows Server 2025 bundles for WorkSpaces Personal and WorkSpaces Core. - Security features include TPM 2.0, UEFI Secure Boot, Credential Guard, HVCI, Secured-core server, and DNS-over-HTTPS. - Existing Windows Server 2016, 2019, and 2022 bundles remain supported. - **AWS Builder ID** - Adds GitHub and Amazon as sign-in options alongside Google and Apple. - Users can access AWS Builder Center, Training and Certification, and Kiro without maintaining separate credentials. - **Amazon Redshift** - Introduces reusable templates for `COPY` operations. - Templates centralize frequently used parameters, improve consistency, and automatically apply future updates to subsequent loads. - The feature is available in commercial and AWS GovCloud Regions. ## Upcoming AWS Events - AWS Summits are scheduled for Paris, London, and Bengaluru. - AWS Community Days are planned in Pune, San Francisco, and Romania. - AWS will participate in NVIDIA GTC 2026 in San Jose. - AWS Community GameDay Europe will offer hands-on troubleshooting challenges across more than 50 cities. For practitioners, the most significant developments are Route 53 Global Resolver for centralized global DNS security, S3 namespaces for organizational naming governance, and AgentCore’s stateful MCP support for more capable AI applications.

Read original(opens in new tab)
Database DesignAmazon S3Aws IamAmazon Redshift+4
cloudflare4 min readCurated summary

From legacy architecture to Cloudflare One

Moving from fragmented VPNs to Cloudflare One is presented as a gradual modernization effort rather than a risky “big bang” cutover. Cloudflare and CDW recommend a tiered, application-aware migration that combines Zero Trust controls with careful dependency analysis and staged deployment. The central conclusion is that legacy applications can gain modern security protections without immediate code rewrites or major downtime. ## Reducing Big-Bang Migration Risk - Large organizations may need to transition hundreds or thousands of applications and users from legacy VPNs. - A single firewall error, dependency failure, or session timeout can disrupt essential services. - These risks often prevent organizations from adopting Zero Trust despite vulnerable, aging infrastructure. - CDW applies lessons from failed deployments to create a risk-aware migration roadmap. - Applications are categorized by complexity, with simpler systems migrated first and legacy systems handled later under tighter controls. - A public-sector migration of 500 applications caused widespread disruption because more than 4,000 applications had not been prioritized or tiered. ## Treating Migration as Application Modernization - Traditional migrations often treat networks as basic connectivity infrastructure and overlook application ecosystems. - CDW analyzes: - Backend databases and APIs - Identity and authentication dependencies - Hidden service-to-service calls - Legacy session behavior - Security requirements are incorporated into the architecture from the beginning rather than added after connectivity is restored. - The migration becomes an application modernization program instead of a simple VPN replacement. ## Protecting Legacy Applications with Cloudflare Access - Cloudflare Access replaces broad network-level VPN access with request-by-request Zero Trust authorization. - Each request can be evaluated using: - User identity - Device posture - Hardware-based MFA - Other contextual signals - This limits lateral movement and reduces the attack surface. - Legacy applications can be “wrapped” with modern security controls without rewriting their code. - Cloudflare Tunnel provides: - An outbound-only connection - SSO and MFA integration - No public IP exposure for the application - Access policies can require endpoint MFA and a device health check before traffic reaches the server. - This approach allows organizations to modernize security incrementally while legacy applications continue operating. ## Pre-Migration Audit ### Architectural and Identity Assessment - Identify whether applications use a federated identity provider such as Okta or legacy local directories. - Map database, API, and backend dependencies. - Verify that hidden API calls and service-token-based Tunnel connections will continue functioning after migration. - Assess whether applying least-privilege controls could break application behavior. ### Establishing a Strategy and Implementation Firebreak - Create separate groups for: - Security strategy and standards - Deployment and operational implementation - This separation prevents deployment speed from overriding requirements designed to limit lateral movement. ### Testing Persistent Sessions - Identify applications that depend on persistent sessions, particularly for users switching between cellular towers. - Cloudflare’s edge architecture and Dynamic Path MTU Discovery (PMTUD) help maintain sessions even when client IP addresses change. - This assessment can identify opportunities to replace rigid legacy hardware with a modern single-pass architecture. ### Categorizing Applications and Setting Timelines - **Tier 0: Modern SaaS applications** - Native SAML/OIDC support - Cloudflare can act as a clientless identity-provider proxy - Estimated effort: 1–3 hours per application - **Tier 1: Internal web applications** - Support identity headers and modern web protocols - Use a clientless reverse proxy with Cloudflare Tunnel - Estimated effort: 3–6 hours per application - **Tier 2: Non-web client-server applications** - Require specific port/protocol support or thick-client configurations - Use both Cloudflare One Client and Cloudflare Tunnel - Estimated effort: 4–8 hours per application A phased migration built around application dependencies, identity readiness, session behavior, and technical complexity offers a safer path to Cloudflare One. Organizations should begin with an audit and pilot, secure legacy applications using Access and Tunnel, and expand tier by tier rather than attempting a single cutover.

Read original(opens in new tab)
Database DesignCloudflareZero TrustSase+4
google3 min readCurated summary

Testing LLMs on superconductivity research questions

LLMs may help physicists navigate complex research, but their reliability depends heavily on the quality and curation of their sources. In a high-temperature superconductivity study, systems grounded in expert-selected literature—especially NotebookLM and a custom retrieval-augmented generation system—outperformed general web-access models. The results suggest that trustworthy scientific AI requires balanced reasoning, strong evidence, and carefully controlled reference collections. ## Evaluating LLMs on Superconductivity - Researchers from Google Research and Cornell University tested whether LLMs could answer expert-level questions in condensed matter physics. - The study focused on cuprate high-temperature superconductors, whose underlying mechanism remains unresolved despite decades of research. - Understanding superconductivity in these materials could help scientists discover compounds that work at higher temperatures. - The field contains thousands of experimental and theoretical papers and competing explanations, making it difficult for researchers—especially newcomers—to establish a balanced view. ## Study Design and Sources - Six systems were evaluated: - GPT-4o - Perplexity - Claude 3.5 - Gemini Advanced Pro 1.5 - Google NotebookLM - A custom retrieval-augmented generation (RAG) system - Four models had broad web access, including 765 open-access experimental papers and 1,553 theoretical papers. - NotebookLM and the custom RAG system used a curated database: - Twelve superconductivity experts selected 15 review articles. - Those reviews contained approximately 3,300 references. - A final collection of 1,726 experimental papers and reviews was assembled. - Experts created 67 difficult questions, including questions about doping levels and evidence for quantum criticality in cuprates. ## Evaluation Criteria Experts used masked reviews and scored responses from 0 to 2 on: - Balance between competing scientific perspectives - Comprehensiveness and factual depth - Conciseness and clarity - Evidence and links to sources - Relevance of supplied images - Qualitative comments ## Results - NotebookLM achieved the strongest overall performance. - The custom RAG system ranked second overall, showing the value of retrieval from the same expert-curated sources. - NotebookLM, Gemini, and the custom RAG system performed best at presenting balanced and comprehensive answers. - NotebookLM provided the strongest evidence and citations but was less concise than the other systems. - Image quality was generally weaker; the custom RAG system performed best among the models that regularly supplied images. - All systems showed areas needing improvement, particularly when addressing nuanced, unresolved research questions. ## Practical Implication For scientific research, LLMs should be paired with expert-curated, quality-controlled literature rather than relying solely on unrestricted web searches. Such systems can serve as research tutors or thought partners, but their answers still require expert verification, especially in fields with competing theories and rapidly evolving evidence.

Read original(opens in new tab)
Database DesignLarge Language ModelsGeminiNotebooklm+2
aws3 min readCurated summary

Twenty years of Amazon S3 and building what’s next | Amazon Web Services

Amazon S3 began in 2006 as a simple web service for storing and retrieving objects, but its emphasis on security, durability, availability, performance, and elasticity enabled it to become foundational infrastructure. Over two decades, it scaled from roughly one petabyte to hundreds of exabytes while preserving API compatibility, reducing prices, and expanding beyond object storage. Amazon’s long-term vision is for S3 to serve as a universal foundation for data, analytics, and AI workloads. ## The Original S3 Philosophy - S3 introduced two basic operations: - `PUT` to store an object - `GET` to retrieve it - The service abstracted away complex infrastructure so developers could focus on applications. - Its five enduring design principles are: - **Security:** Data is protected by default. - **Durability:** Designed for 11 nines of durability, with a lossless operating model. - **Availability:** Failure is assumed and handled throughout the system. - **Performance:** Storage capacity can grow without degrading performance. - **Elasticity:** Capacity expands and contracts automatically. ## From One Petabyte to Hundreds of Exabytes - At launch, S3 had approximately: - One petabyte of capacity - 400 storage nodes across 15 racks and three data centers - 15 Gbps of bandwidth - A maximum object size of 5 GB - A price of $0.15 per GB - Today, S3: - Stores more than 500 trillion objects. - Serves over 200 million requests per second. - Operates across 123 Availability Zones in 39 AWS Regions. - Supports objects up to 50 TB—10,000 times larger than the original limit. - Storage prices have fallen by roughly 85%, to slightly above 2 cents per GB. - S3 Intelligent-Tiering has saved customers more than $6 billion in storage costs. - The S3 API has become an industry standard, with many other storage systems offering compatible interfaces. ## Backward Compatibility and Long-Term Reliability - Code written against S3 in 2006 still works without modification. - AWS has repeatedly replaced disks, storage systems, and request-processing code while preserving access to older data. - This compatibility reflects S3’s goal of remaining infrastructure that “just works” despite continuous internal change. ## Engineering for Durability and Scale - Microservices continuously inspect every byte across the fleet. - Auditor services detect degradation and automatically trigger repair and re-replication. - Automated formal methods mathematically verify correctness in areas such as: - The index subsystem - Cross-Region replication - Access policies - AWS has progressively rewritten performance-critical components in Rust over the past eight years. - Rust improves performance while preventing memory-safety bugs and other classes of errors at compile time. - S3 follows the principle that scale should improve the service: larger, more distributed workloads become increasingly decorrelated, improving reliability for all customers. ## S3 as a Foundation for Data and AI Amazon’s future vision is for customers to store data once in S3 and work with it directly, avoiding costly copies and specialized systems. - **S3 Tables** provides managed Apache Iceberg tables with automated maintenance to improve query performance and reduce storage costs. - **S3 Vectors** supports semantic search and retrieval-augmented generation, with up to 2 billion vectors per index and sub-100 ms query latency. - Within five months of launch, customers created over 250,000 indexes, ingested more than 40 billion vectors, and executed over 1 billion queries. - **S3 Metadata** enables centralized, faster data discovery without recursively listing large buckets. These additions extend S3 from inexpensive object storage into a broader platform for analytics, search, and AI while retaining its scale and cost advantages.

Read original(opens in new tab)
Database DesignRustAPIMicroservices+4
kakao4 min readCurated summary

From Student to Developer: Learning Rational Choices Over Right Answers—From DB and Security to AI

The onboarding of 40 new Kakao developers shifted their perspective from making features work to designing systems that survive real-world operations. Across databases, security, and AI, they learned that there is rarely one perfect answer; the best choice depends on scale, risk, maintainability, and business needs. The central lesson was to replace theoretical correctness with responsible, adaptable engineering judgment. ## Database: From Finding the Right Answer to Preparing for Change - Database design must be evaluated by whether it can withstand traffic, schema changes, and operational demands—not only by theoretical correctness. - Foreign keys are not automatically the best choice: - They can introduce locking, performance, and flexibility concerns. - Referential integrity can instead be managed at the application layer, provided testing and correction processes are strong. - Soft deletion, using fields such as `deleted_at`, supports auditability and recovery and is often an essential operational strategy. - Indexes should be selected according to the questions the database must answer: - B-tree, GIN, GiST, SP-GiST, and vector indexes serve different data and query patterns. - Execution plans reveal whether SQL uses indexes or performs full table scans, directly affecting I/O and response times. - Duplication is not always harmful: - Intentional denormalization can avoid expensive joins. - Snapshot data can simplify reads and preserve the information needed by a business workflow. - In MongoDB, embedding selected related data can make screen queries much simpler than relying exclusively on references. - Different database systems embody different trade-offs among performance, consistency, scalability, and operational cost. - The training covered MySQL high availability, PostgreSQL primary-key structures, cloud-native systems such as Neon, and the broader storage-to-analysis pipeline of Hadoop and Spark. - The resulting mindset favors designs that are safe to change and affordable to operate over designs that are theoretically perfect. ## Security and IT: From Someone Else’s Responsibility to a Personal Default - Security became a direct consequence of developers’ code rather than merely a compliance or infrastructure concern. - Everyday safeguards such as development/production separation, VPNs, and antivirus software demonstrate that safety often requires accepting some inconvenience. - DDoS defense is not only about blocking traffic: - It can be difficult to distinguish an attack from legitimate traffic spikes caused by a popular event. - Developers should apply basic controls such as rate limiting and escalate suspicious activity through established response channels. - Hands-on API exploitation made vulnerabilities concrete and encouraged developers to view security through an attacker’s perspective. - Security must be continuous: - AI is increasingly being used both to discover vulnerabilities and to strengthen attacks. - Social-engineering methods involving QR codes, app permissions, and human behavior require more than purely technical defenses. - Security checks should be integrated from the beginning of development, not performed only at the end. - Software quality also depends on people: - Code should remain understandable enough for another developer to take over quickly. - Strong engineering means choosing and communicating the most appropriate solution for the business context, not merely finding a technically possible one. ## AI: From Chatting with Models to Designing Systems - An AI agent is not simply a model; it is an architecture composed of tools, routing logic, error handling, and model calls. - Agent development applies familiar software-engineering practices to probabilistic models. - Because LLM outputs can vary, reliable systems need deliberate controls: - Prompt chaining breaks large tasks into smaller steps and limits context contamination. - Few-shot examples clarify required output formats. - Routing selects different prompts or workflows based on conditions. - Multi-agent systems divide responsibilities among specialized agents, echoing the modularity and scalability principles of microservices. - RAG reduces hallucinations structurally by: - Chunking documents. - Searching for semantically similar vectors. - Supplying retrieved information to the model as additional context. - MCP exposes internal systems and data as callable tools, effectively enabling remote function calling and connecting AI to enterprise capabilities. - Effective AI use shifted from criticizing poor answers to specifying clear objectives, formats, examples, context, and supporting data. - The goal is not merely to receive an intelligent response, but to design a system that consistently produces intelligent behavior. The training ultimately marked a transition from student-style problem solving to professional engineering. Developers should consider operational resilience, security, maintainability, and business value, then make and clearly explain the most reasonable choice for the circumstances.

Read original(opens in new tab)
Database DesignLarge Language ModelsAI AgentsPostgreSQL+4