<h2>Must know</h2>
<ul>
<li><strong>Microsoft patches zero-days and AI vulnerabilities</strong>: Microsoft’s February 2026 Patch Tuesday addressed 59 vulnerabilities, including six actively exploited zero-days. Critically, it also patched remote code execution flaws in GitHub Copilot and multiple IDEs (VS Code, Visual Studio, JetBrains) stemming from AI prompt injection vulnerabilities. These AI-related flaws pose significant risks to developers due to potential access to sensitive credentials and infrastructure keys, as reported by Krebs on Security. <a href="https://app.daily.dev/feed-by-ids?id=zk4Rm0PBv&amp;id=BOPPPXswR&amp;id=i2YjgAdJA&amp;id=oUCth2teJ&amp;id=uxvhPzT2r&amp;id=zFYu8Wov8" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents introduce new security risks</strong>: The rapid deployment of autonomous AI agents is creating significant security vulnerabilities, with experts warning about supply chain attacks, hidden commands, and hallucinated commands in agent ‘skills.’ GitGuardian, which just raised $50M, is focusing on securing these non-human identities, while researchers have demonstrated command hijacking attacks against embodied AI systems like robots. This highlights a critical gap between AI deployment speed and the maturity of its security frameworks. <a href="https://app.daily.dev/feed-by-ids?id=0rQwEgnbB&amp;id=QRBfTNO1G&amp;id=ZVqULJLz3&amp;id=gUwIhEnZO&amp;id=l8YCDT7qf&amp;id=mmQEWnSC4&amp;id=BOPPPXswR&amp;id=i2YjgAdJA&amp;id=oUCth2teJ&amp;id=uxvhPzT2r" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Open-source GLM-5 model claims parity with proprietary AI</strong>: <a href="http://Z.ai" target="_blank" rel="noopener nofollow">Z.ai</a> has released GLM-5, a new open-source large language model under an MIT License, claiming it matches proprietary models like Claude Opus 4.6 and GPT 5.3 Codex for long-horizon coding tasks. With 744B parameters and novel asynchronous RL infrastructure, it aims for best-in-class performance among open-source models on reasoning, coding, and agentic tasks. Modal is offering free API access to GLM-5 until the end of April. <a href="https://app.daily.dev/posts/bOspGaOnW" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>VS Code enhances AI agent capabilities</strong>: VS Code 1.109 introduces AI-powered features like agent clarifying questions, MCP apps for interactive UIs, default-enabled agent skills, and parallel sub-agent execution, alongside expanded AI model options through GitHub Copilot. <a href="https://app.daily.dev/posts/QJP16aVAU" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Databricks launches Lakeflow for data engineering</strong>: Databricks announced Lakeflow, a unified data engineering solution on Azure Databricks for ingestion, transformation, orchestration, and governance, featuring AI-powered debugging and serverless compute. <a href="https://app.daily.dev/feed-by-ids?id=Bf1dK14FP&amp;id=DckO36Vaq&amp;id=cW6KgKgEK&amp;id=omSTWjEfj" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Kubernetes is now dominant for AI workloads</strong>: Kubernetes is now the infrastructure of choice for 82% of production AI workloads, but cultural barriers like siloed teams are the biggest hurdle to scaling AI initiatives, according to a CNCF report. <a href="https://app.daily.dev/posts/MrQz3g4zd" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Erlang/OTP 29.0-rc1 released with new features</strong>: Erlang/OTP 29.0-rc1 introduces native records, multi-valued comprehensions, a new <code>is_integer/3</code> guard BIF, and defaults to quantum-resistant key exchange for SSH, while removing 32-bit Windows builds. <a href="https://app.daily.dev/posts/fTCktIrzF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New web standard for AI agent interaction proposed</strong>: WebMCP, a proposed web standard, aims to enable websites to expose structured tools directly to in-browser AI agents, replacing brittle DOM scraping with explicit function calls and declarative HTML annotations. <a href="https://app.daily.dev/feed-by-ids?id=Ib422mkez&amp;id=NpNZ22j1b&amp;id=R0k8YNGTx" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Rust web framework Leptos gains traction</strong>: Leptos, a Rust-based web framework created by an Episcopal priest, is gaining significant community adoption with over 19,000 GitHub stars, demonstrating performance comparable to Solid and faster than many other frameworks. <a href="https://app.daily.dev/feed-by-ids?id=cCyxGk3Tq&amp;id=gkCUaPXMf" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Amazon OpenSearch Serverless introduces Collection Groups</strong>: Amazon OpenSearch Serverless now supports Collection Groups, allowing shared compute resources across multiple collections to reduce costs while maintaining security isolation, particularly useful for multi-tenant workloads. <a href="https://app.daily.dev/posts/5AicsK9Gn" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI will make you obsolete, unless you buy our tools</strong>: Apparently, AI models have reached an ‘inflection point’ where they can complete complex tasks autonomously, leading to 50% of entry-level white-collar jobs being eliminated in 1-5 years. The solution? Immediately adopt paid AI tools and experiment daily, because the ‘window for competitive advantage’ is closing. So, if you’re not buying into the hype and the tools, you’re basically signing your own pink slip, according to some ‘industry leaders.’ <a href="https://app.daily.dev/feed-by-ids?id=jQIyW7C3o&amp;id=lGyFb7XTB&amp;id=waJpUtc5V" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-powered testing is the death of traditional testing</strong>: Datadog and GitHub Copilot are pushing new AI-powered testing tools, with one article boldly proclaiming ‘The Death of Traditional Testing.’ These tools promise to automatically generate tests, infer code change intent, and ‘auto-heal’ tests. Because, you know, humans maintaining tests is just so passé. Soon, we’ll all be sitting back while AI writes, tests, and probably deploys our code, leaving us to ponder our existential purpose. <a href="https://app.daily.dev/feed-by-ids?id=gZJqHPom7&amp;id=j254S16VC&amp;id=sJ5P2kSW8" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Orbital AI data centers are the future (if money grows on trees)</strong>: Some folks are seriously pushing for orbital AI data centers, with SpaceX and Google apparently on board. The only minor hurdles? A 1 GW space facility costs nearly three times its ground equivalent, requiring launch costs to drop from $3,600/kg to $200/kg, cheaper satellites, and solutions for thermal management and radiation hardening. Oh, and current economics only favor inference, not training. But hey, if you can just ignore basic physics and economics, it’s a brilliant idea! <a href="https://app.daily.dev/feed-by-ids?id=0lLQRkBLb&amp;id=Zm3R4GzeA&amp;id=bYavSTao6&amp;id=daHpBJ83O&amp;id=j9QS83E6P" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
