エピソード

  • MongoDB hacked, Xcode Cloud included with the Apple Developer Program. Fly.io has Managed Postgres, and Microsoft Office in VR
    2023/12/19

    MongoDB Security Incident: MongoDB reported unauthorized access to some of its systems. Notably, this incident didn't affect their hosted service, MongoDB Atlas. However, they advised users to change their passwords, leading to temporary unavailability of Atlas during the rush for password updates. Compromised data may include names, phone numbers, and email addresses. Users are urged to update their MongoDB passwords as a precaution.

    LinkedIn's Cloud Strategy Shift: LinkedIn has shelved its plans to transition to Microsoft Azure, citing challenges in migrating legacy code deeply integrated with their existing infrastructure. While still utilizing Azure for some components like content delivery, LinkedIn doesn't see sufficient ROI in a full-scale migration. This represents a setback for Microsoft in its bid to rival Amazon AWS.

    Apple Developer Program Enhancements: The Apple Developer Program now includes 25 hours of Xcode Cloud access. Xcode Cloud facilitates cloud-based app building, particularly useful for App Store submissions, with integrated TestFlight and App Store tools. This addition, effective January, encourages developers to explore Xcode Cloud's capabilities.

    Fly.io Partners with Superbase for Managed Postgres: Fly.io announced its managed Postgres service, powered by Superbase. This service simplifies database management by handling scaling, backups, and upgrades, and offers direct Postgres access along with GraphQL and REST API features.

    Microsoft Office in VR: MetaQuest users can now use Microsoft Office apps like Excel, Word, and PowerPoint in VR, available in the MetaQuest store. While other Office 365 apps like Outlook and Teams are expected to follow, the current VR versions might require a Bluetooth keyboard for optimal functionality.

    Links:

    https://www.mongodb.com/alerts

    https://developer.apple.com/news/?id=ik9z4ll6

    https://supabase.com/blog/postgres-on-fly-by-supabase

    https://www.engadget.com/microsoft-office-apps-arrive-on-meta-quest-vr-headsets-123030297.html

    続きを読む 一部表示
    4 分
  • Apps still vulnerable to log4shell exploit, Deno 1. 39 realeased, Programming language rankings delayed, WebP compression introduces artifacts, Zero downtime PG upgrades
    2023/12/14

    Daily news roundup in Tech.

    Links:

    https://www.theregister.com/2023/12/11/log4j_vulnerabilities/

    https://redmonk.com/rstephens/2023/12/14/language-rankings-update/

    https://deno.com/blog/v1.39

    https://eng.aurelienpierre.com/2021/10/webp-is-so-great-except-its-not/

    https://knock.app/blog/zero-downtime-postgres-upgrades

    続きを読む 一部表示
    6 分
  • EU AI Law, Minstral AI releases Mixtreal 7bx8, Safari 17. 2 released, Slack reports working overtime makes you less productive
    2023/12/12
    • EU AI Regulation: Landmark rules finalized; effective 2025. Bans on certain AI uses, with a focus on ethical application in law enforcement and public services.
    • Minstrel AI Mixtrol 8x7B: New AI model released, outperforming GPT-3.5 in benchmarks. Details on upcoming models and EU market implications.
    • Safari 17.2 Release: Features 39 enhancements and 169 fixes, including HTML accordions, one-time codes, and responsive image support.
    • Slack's Productivity Survey: Reveals a 20% productivity boost in employees working regular hours versus overtime. Highlights the risks of skipping breaks on employee burnout.


    続きを読む 一部表示
    5 分
  • Auth0 increasing prices, Google admits Gemini demos were not real, OpenAI acknowledges ChatGPT is getting lazy
    2023/12/09

    In our latest episode, we dive into some striking tech developments. I kick things off with the shocking news of Auth0 ramping up its prices by 300%. This substantial increase, especially for a service already known as one of the priciest, leads me to ponder if alternatives like Firebase or Superbase might be more viable. T

    hen, I tackle the revelation about Google's Gemini AI demo. They've admitted that the demo was, in parts, exaggerated, which leaves me questioning the real capabilities of the Gemini AI upon its launch.

    Lastly, I share some light-hearted insights on OpenAI's ChatGPT. There's been chatter that the AI seems less responsive or 'lazy' as we approach the holiday season. Some users, including a friend of mine, even experimented with promising 'tips' to ChatGPT for better responses, with surprisingly effective results.

    続きを読む 一部表示
    4 分
  • Governments spying on users via push notifications, Sqlite announce JSONB support, better serverless pattern FLAME by Fly.io
    2023/12/07

    Apple and Google acknowledge government surveillance through push notifications, raising privacy concerns.

    SQLite enhances its offerings with JSONB support, a boon for developers using SQLite and JSON.

    Additionally, we delve into Fly.io's innovative approach to serverless computing with the FLAME pattern, providing a cost-effective, efficient solution for handling high CPU workloads in applications.

    Links:

    https://www.androidauthority.com/apple-google-push-notification-surveillance-3392252/

    https://sqlite.org/forum/forumpost/fa6f64e3dc1a5d97

    https://fly.io/blog/rethinking-serverless-with-flame/

    続きを読む 一部表示
    5 分
  • DeepMind announce Gemini, Sentry release Spotlight, Imagine with Meta, Wing a programming language for the cloud
    2023/12/06

    This episode starts with Google DeepMind's Gemini, a versatile language model with impressive speed and capabilities. SpotlightJS by Sentry, a new local debugging tool, also catches my eye for its cross-language support. Then, we touch on Meta's Imagine, an intriguing new image generator. Finally, we discuss Wing, a programming language that simplifies cloud development.

    Links:

    • https://deepmind.google/technologies/gemini/#introduction
    • https://spotlightjs.com/
    • https://techcrunch.com/2023/12/06/meta-launches-a-standalone-ai-powered-image-generator/
    • https://www.winglang.io/


    続きを読む 一部表示
    6 分
  • Astro 4.0 launch, Signal's (Electron) security issue, 23andMe breach, and finetuning AI for Magic the Gathering
    2023/12/06

    In this episode, we dive into the latest tech updates: the launch of Astro 4.0, enhancing web development with faster builds and new features; a newly discovered security issue in Signal's Electron app, underscoring the need for robust software security; the massive data breach at 23andMe affecting 6.9 million users, highlighting the importance of secure APIs and two-factor authentication; and an intriguing exploration of fine-tuning the Minstrel 7B AI model for selecting Magic the Gathering cards, demonstrating the rapid advancements and challenges in the AI field.

    Links:

    https://astro.build/blog/astro-4/

    https://blodeuweddlabs.com/news/signal-desktop-ui-redressing-issue/

    https://techcrunch.com/2023/12/04/23andme-confirms-hackers-stole-ancestry-data-on-6-9-million-users/

    https://generallyintelligent.substack.com/p/fine-tuning-mistral-7b-on-magic-the


    続きを読む 一部表示
    5 分
  • Railsway moves from GCP after downtime, Django 5.0 out now, DeepMind tests ChatGPT's limits while Unsloth AI launches a fast tool for tuning language models.
    2023/12/05

    This episode delves into the latest developments shaking up the tech world. We begin by looking at Railsway's recent downtime troubles and their decision to transition away from the Google Cloud Platform. Discover what this means for the future of cloud computing and how it reflects a growing trend towards independent infrastructure.

    Next up, we celebrate the release of Django 5.0. We'll explore its new features, including enhanced support for Python versions 3.10 to 3.12, improved admin interface, simplified templating for field groups, and the introduction of database computed default values. This segment is a must-listen for web developers and Python enthusiasts.

    We then dive into DeepMind's intriguing experiment with ChatGPT, uncovering how they managed to extract training data from the AI model. This segment sheds light on AI development's potential vulnerabilities and ethical considerations.

    Finally, we spotlight Unsloth AI's groundbreaking tool for fine-tuning large language models. Learn how this tool, available in both open-source and paid versions, accelerates the fine-tuning process by 80%, reduces memory usage by 50%, and retains full accuracy.

    続きを読む 一部表示
    4 分