Best of Cloud — September 2025

  1. 1
    Article
    Avatar of modal_labsModal·37w

    Introducing Notebooks

    Modal launches Notebooks, a collaborative cloud-based computing environment that provides instant GPU-enabled Python kernels starting in under 5 seconds. The platform offers real-time collaborative editing, automatic resource scaling from 0.125 CPUs to 8 H100/B200 GPUs, and seamless integration with Modal's existing infrastructure including Volumes and Functions. Key features include fast cold-start times, automatic idle shutdown to reduce costs, shared environments across teams, and modern development tools like LSP support and AI completions.

  2. 2
    Video
    Avatar of codeheadCodeHead·37w

    Should YOU Become a Backend Dev?

    Backend development in 2025 involves managing complex microservices architectures, cloud infrastructure, AI integrations, and security concerns. While the core responsibilities remain building APIs and managing databases, developers now juggle dozens of services, navigate extensive cloud platforms like AWS, and handle AI model integrations. The role has expanded to include security engineering responsibilities, with constant focus on preventing data breaches. Despite the challenges of debugging distributed systems and handling 3 AM alerts, backend developers remain essential for keeping internet infrastructure running smoothly.

  3. 3
    Article
    Avatar of gcpGoogle Cloud·37w

    Now available: Rust SDK for Google Cloud

    Google Cloud has released an official Rust SDK providing access to over 140 Google Cloud APIs. The SDK includes built-in authentication support, comprehensive documentation, and covers services like Vertex AI, Cloud Key Management, and Identity Access Management. Previously, developers had to rely on unofficial SDKs with limited features and security concerns. The new SDK is available on crates.io and GitHub, offering a supported and idiomatic way to integrate Rust applications with Google Cloud services.

  4. 4
    Article
    Avatar of platformaticPlatformatic·35w

    Node.js on Kubernetes: Unveiling Myths and Costs

    Running Node.js on Kubernetes presents significant challenges due to fundamental mismatches between Node.js's lightweight, event-driven architecture and Kubernetes' resource allocation model. Common myths include believing autoscaling works seamlessly out-of-the-box, when in reality scaling delays can cause performance issues during traffic spikes. The rigid CPU/memory request/limit system forces teams to choose between costly overprovisioning or risky underprovisioning. To optimize Node.js in Kubernetes, teams should use smarter scaling signals like event loop lag, implement finer-grained resource strategies, reduce scaling reaction times, treat cost as a first-class metric, and recognize that Node.js requires different management approaches than traditional JVM applications.

  5. 5
    Article
    Avatar of metalbearMetalBear·36w

    MetalBear raises $12.5m to make cloud software development 100x faster

    MetalBear secured $12.5M in seed funding led by TLV Partners to advance mirrord, a development tool that enables developers to test cloud applications locally in seconds rather than hours. The tool works by intercepting system calls at the C standard library level, allowing local code to run with the context and dependencies of production cloud environments. Companies like SurveyMonkey and zooplus report significant productivity gains, with some seeing 50% faster time-to-ship and 20% developer time savings. The fully remote team of 25 people across 15 countries plans to expand mirrord's capabilities and ease of adoption.

  6. 6
    Article
    Avatar of techworld-with-milanTech World With Milan·35w

    Scale from zero to million users on Azure

    A comprehensive guide to scaling applications from single users to millions on Azure, covering architectural evolution through six stages. Explores key concepts including stateless design, caching strategies, load balancing, database scaling, microservices decomposition, and multi-region deployment. Details Azure-specific services like App Service, SQL Database, Redis Cache, Front Door, and Cosmos DB for each scaling phase, with practical trade-offs between consistency and availability at global scale.

  7. 7
    Article
    Avatar of infoqInfoQ·36w

    Vercel Introduces AI Gateway for Multi-Model Integration

    Vercel has launched AI Gateway for production use, offering a unified API endpoint to access multiple AI model providers like OpenAI, Anthropic, and Google. The service features sub-20ms latency routing, automatic failover between providers, bring-your-own-key authentication without token markup, and comprehensive observability tools. It competes with services like OpenRouter by focusing on low-latency performance and integration with Vercel's developer ecosystem.

  8. 8
    Video
    Avatar of techwithlucyTech With Lucy·36w

    The Future of Databases is here... (What you need to know)

    Modern databases are evolving to handle both transactional and analytical workloads in real-time, eliminating the need for complex multi-database architectures. Traditional setups require separate systems for applications and analytics, causing data delays and increased complexity. New unified database platforms can process transactions, run analytics, and support AI workloads simultaneously, enabling instant dashboards, real-time fraud detection, and live inventory systems. The future points toward databases that natively support vector search, unstructured data, and AI workloads without requiring multiple tools or complex data pipelines.

  9. 9
    Article
    Avatar of techcentralTechCentral·37w

    OpenAI’s staggering burn rate: $115-billion by 2029

    OpenAI has dramatically increased its projected spending to $115 billion through 2029, up $80 billion from previous estimates, as it scales infrastructure for ChatGPT and AI services. The company plans to develop custom chips with Broadcom and build its own data centers to control soaring cloud computing costs, with annual burn rates expected to reach $45 billion by 2028.

  10. 10
    Article
    Avatar of tcTechCrunch·35w

    It isn’t your imagination; Google Cloud is flooding the zone

    Google Cloud is pursuing a strategy to capture emerging AI startups while industry giants like Nvidia, OpenAI, Microsoft, and Amazon form massive exclusive partnerships worth hundreds of billions. Google's COO Francis deSouza reveals that 60% of generative AI startups use Google Cloud, with the company offering $350,000 in credits, technical support, and open infrastructure access. This approach contrasts with competitors' mega-deals, positioning Google to benefit from the next wave of AI unicorns before they become too expensive to court.

  11. 11
    Article
    Avatar of clickhouseClickHouse·34w

    How we scaled raw GROUP BY to 100 B+ rows in under a second

    ClickHouse Cloud introduces parallel replicas, enabling GROUP BY queries to scale across thousands of cores automatically. The feature processes 100 billion rows in under half a second without pre-aggregation or data reshuffling. By distributing work across multiple nodes using partial aggregation states, queries achieve terabyte-per-second throughput with linear scaling. The system uses granule-based load balancing and safeguards to optimize performance, making analytical queries interactive even at massive scale.

  12. 12
    Article
    Avatar of thedailywtfThe Daily WTF·35w

    One Last ID

    A web development company made a costly SQL mistake by adding an unnecessary FROM clause to a last_insert_id() query, causing MySQL to return millions of duplicate values instead of a single ID. This simple error resulted in 30MB responses for each INSERT operation, generating terabytes of unexpected data transfer costs and blowing the operational budget. The technical fix was simple, but coordinating the deployment took six weeks, including two weeks just to convince the company the problem existed.

See all Cloud archives