<h2>Must know</h2>
<ul>
<li><strong>Google’s Gemini faces IP theft and nation-state exploitation</strong>: Google blocked over 100,000 attempts to extract proprietary reasoning capabilities from its Gemini AI model, which it considers intellectual property theft. Simultaneously, Google’s threat report revealed nation-state actors are integrating Gemini into attack operations for social engineering, vulnerability analysis, and phishing, with new malware families like HONESTCUE embedding Gemini’s API directly to generate and execute malicious code. <a href="https://app.daily.dev/posts/OQLlwLpSw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents create new enterprise security gaps</strong>: AI agents are creating significant governance gaps in enterprise systems by acting as autonomous identities with access to sensitive data, which traditional IAM/PAM tools can’t handle. A unified control plane is needed to manage AI as both an identity and data problem, enforcing least privilege and data classification across human and non-human actors, especially with the rise of the Model Context Protocol. <a href="https://app.daily.dev/posts/7qf6CcnTl" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux 7.0 integrates post-quantum cryptography</strong>: The Linux 7.0 kernel is integrating ML-DSA, a quantum-resistant signature algorithm, for module signing, replacing SHA-1 support. This move is part of a broader migration to post-quantum cryptography, standardized by NIST, to protect against future quantum attacks that threaten current RSA and ECDSA algorithms. This is a significant step in securing critical infrastructure against emerging threats. <a href="https://app.daily.dev/feed-by-ids?id=6DJe6t3uU&amp;id=Ukr75F53X&amp;id=Vay0ocYOF&amp;id=Vh5TboDqs&amp;id=m5L4rWAXQ&amp;id=ppwBwOuN5&amp;id=uGdXnDZH4&amp;id=xVcPYGmKq" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Minimax launches M2.5 AI model</strong>: Minimax launched its M2.5 frontier AI model, claiming it’s over 90% cheaper than Claude Opus or GPT-5 models and excels at software engineering tasks. <a href="https://app.daily.dev/posts/VQt2ETdNG" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>TypeScript 6.0 beta is the last JavaScript version</strong>: TypeScript 6.0 beta is the final version built on JavaScript, with TypeScript 7.0 transitioning to a Go-based compiler for improved performance. <a href="https://app.daily.dev/posts/62kNn8zlZ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Go 1.26 released with Green Tea GC</strong>: Go 1.26 was released with the Green Tea garbage collector enabled by default, a new <code>new</code> syntax for initial values, and experimental SIMD operations. <a href="https://app.daily.dev/feed-by-ids?id=1ObNpwyIZ&amp;id=mj2oUwkea" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI launches GPT-5.3-Codex-Spark on Cerebras</strong>: OpenAI released GPT-5.3-Codex-Spark, a coding model running on Cerebras CS3 accelerators, prioritizing speed with 15 times faster inference than previous models. <a href="https://app.daily.dev/feed-by-ids?id=2DKu3o06c&amp;id=2G5oHg0oJ&amp;id=A9s0KznyU&amp;id=CacFMCP2L&amp;id=PFJ429xnc&amp;id=TgVTzT7ZG&amp;id=dqsxZt8l5&amp;id=f8bwmVFOm&amp;id=iVwXrkdSj&amp;id=yYXkpJGbb" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>HBM4 memory enters mass production</strong>: Samsung and Micron have begun mass production and shipping of HBM4 memory, the next-generation high-bandwidth memory for AI accelerators, with Micron’s 2026 output already pre-sold. <a href="https://app.daily.dev/feed-by-ids?id=OUBYQ2PHL&amp;id=gpkg8awYO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Orchestra: a new self-hosted PaaS</strong>: Orchestra, an open-source, self-hosted PaaS, was released for managing physical servers and bare metal infrastructure, automating provisioning and application deployment without agents. <a href="https://app.daily.dev/feed-by-ids?id=0nUdXf1b9&amp;id=C2t5MyKnp&amp;id=jNBhBoH4V&amp;id=nwGZvk6OS&amp;id=sjdf7M4Ia" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cloudflare open-sources Rust library for zero-downtime upgrades</strong>: Cloudflare open-sourced ecdysis, a Rust library enabling zero-downtime service upgrades through graceful process restarts, proven in production for five years. <a href="https://app.daily.dev/posts/tdBY3TVs4" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Spotify’s top devs now code from their phones</strong>: Spotify’s top developers reportedly haven’t written code manually since December, using Claude Code to ship features and fix bugs from their phones and Slack. Because, you know, real engineering happens on a phone now. I’m sure the architectural integrity is just <em>chef’s kiss</em>. <a href="https://app.daily.dev/feed-by-ids?id=Sq7jBWt9y&amp;id=s5BDVI2HR" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic’s AI-generated C compiler is mostly hype</strong>: Anthropic’s AI-generated C compiler, built by 16 Claude Opus agents in two weeks, is being touted as a breakthrough. Turns out, it fails basic tests and relies on existing tools. It’s almost like AI still needs humans to do the actual hard work, but hey, great marketing! <a href="https://app.daily.dev/posts/mZHhw9XzP" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI code generation: more speed, more privilege escalation</strong>: AI adoption is intensifying the tension between development speed and security, with AI-generated code showing a 322% increase in privilege-escalation risks. But don’t worry, declarative security policies are here to save the day! Because adding more layers of abstraction always solves the underlying problem, right? <a href="https://app.daily.dev/posts/1ssjvObXV" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
