<h2>Must know</h2>
<ul>
<li><strong>Critical open-source infrastructure needs funding</strong>: The sole maintainer of the critical <code>sudo</code> utility for over 30 years, Todd C. Miller, is seeking sponsorship after Quest Software ended its support. This highlights the precarious state of essential open-source infrastructure. Meanwhile, <code>sudo-rs</code>, a Rust-based memory-safe alternative, is gaining traction, with Ubuntu 25.10 planning to adopt it as the default <code>sudo</code> utility by October 2025. <a href="https://app.daily.dev/posts/E670NHqPD" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agent security is still a dumpster fire</strong>: AI agent security is a mess, with 1.5 million out of 3 million agents in large US and UK enterprises operating without oversight, leading to security incidents like data leaks and unauthorized actions. Researchers found critical vulnerabilities in n8n’s AI automation platform, allowing remote code execution, and warn that AI coding tools accelerate exploit development, enabling cybercriminals to reverse-engineer exploits in hours. This is a critical and evolving security challenge, as reported by JFrog and Gravitee. <a href="https://app.daily.dev/feed-by-ids?id=ycPDVy6OP&amp;id=Hv04ODVQw&amp;id=ehksC8VMf&amp;id=gy7WeHs5A&amp;id=kssiMjacP&amp;id=tpe4bjIum" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Deno Deploy goes GA with secure sandbox for untrusted code</strong>: Deno Deploy is now generally available, offering zero-config deployment for JavaScript/TypeScript apps, Node.js framework support, and integrated serverless PostgreSQL. They also launched Deno Sandbox in beta, providing instant Linux microVMs for securely running untrusted code, especially LLM-generated code, with hypervisor-level isolation and sub-second boot times. This is a significant step for secure execution environments for AI-generated code. <a href="https://app.daily.dev/posts/Jmdw76tdu" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Vercel v0 pivots to enterprise AI development</strong>: Vercel rebuilt its v0 AI-powered development platform, evolving it into an enterprise-grade AI coding platform with security features, integrations with Snowflake and AWS, and Git-based workflows for production-ready code. <a href="https://app.daily.dev/posts/8wRm1h8NM" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Databricks Lakebase is now generally available</strong>: Databricks Lakebase, a new operational database architecture, is now generally available on AWS, offering a serverless Postgres service that scales automatically and runs directly on data lakes for unified operational, analytical, and AI workloads. <a href="https://app.daily.dev/posts/qtDOznoeU" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Snowflake expands AI data cloud with new offerings</strong>: Snowflake launched Cortex AI, a serverless generative AI suite, and a PostgreSQL database-as-a-service within its AI Data Cloud, integrating transactional workloads with analytics and AI without data movement. <a href="https://app.daily.dev/feed-by-ids?id=1F9HXgJiK&amp;id=ppuh0CVeq&amp;id=qgdGrzKKG" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI details Codex CLI internals</strong>: OpenAI released the first part of a series detailing the internal design of Codex, their software development agent, focusing on its agent loop architecture and challenges like managing LLM inference performance. <a href="https://app.daily.dev/feed-by-ids?id=DWHoaTisZ&amp;id=Dick5lxkX&amp;id=JSYEwKNu7&amp;id=YQpwOBE0D&amp;id=aSHfIm3oS&amp;id=cBoJTA7Xz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AWS IAM Identity Center gets multi-region replication</strong>: AWS IAM Identity Center now supports multi-Region replication for workforce identities and permission sets, enhancing resiliency and enabling applications closer to users while maintaining centralized configuration. <a href="https://app.daily.dev/posts/wVGc4PkCH" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Dependabot enhances security with OIDC support</strong>: Dependabot now supports OpenID Connect (OIDC) authentication for private registries, allowing it to dynamically obtain short-lived credentials from cloud identity providers, reducing attack surface and simplifying credential management. <a href="https://app.daily.dev/feed-by-ids?id=P7dsW78XN&amp;id=YmYpo2Y1G&amp;id=iDGQonMWb" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GNU Hurd project shows significant progress</strong>: GNU Hurd, the microkernel-based Unix replacement, is showing significant progress after 36 years, achieving 80% Debian package compilation, supporting major desktop environments, and gaining upstream support from gcc, glibc, LLVM, and Rust. <a href="https://app.daily.dev/posts/XjM52iBH8" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>X replaces search with Grok LLM</strong>: X (formerly Twitter) has apparently replaced its entire search functionality with Grok, their LLM, because fixing the existing search system was “too difficult.” So, instead of engineering a robust search, they just threw billions of dollars at GPUs and an LLM. What could possibly go wrong with a search engine that hallucinates? <a href="https://app.daily.dev/posts/1TELuPW5m" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Take-Two says GTA 6 is handcrafted, not AI-generated</strong>: Take-Two Interactive CEO Strauss Zelnick proudly declared that generative AI plays “zero part” in GTA 6’s development, emphasizing its “handcrafted design.” This comes after their stock dropped when Google announced Project Genie, an AI tool for creating playable game worlds. It’s almost like they’re trying to reassure investors that their multi-billion dollar franchise isn’t going to be replaced by a prompt, or that their artists aren’t just glorified prompt engineers. <a href="https://app.daily.dev/posts/chwMGhSO9" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Polish railway portal randomizes schedules for mobile users</strong>: A Polish railway portal is displaying randomized train schedules to mobile users based on their IP addresses, with departure times varying by ±60 minutes. PKP, the railway company, declined to comment, citing “security reasons.” Because nothing says “secure” like intentionally providing incorrect information to paying customers. I’m sure this will build immense trust in their digital services. <a href="https://app.daily.dev/feed-by-ids?id=NpQixwGEK&amp;id=hQA36mWZO" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
