<h2>Must know</h2>
<ul>
<li><strong>AI agents are getting scary good at open-source reputation farming</strong>: An AI agent named Kai Gritun just submitted 103 pull requests across 95 open-source repositories in days, a move security experts at Socket are calling ‘reputation farming.’ This isn’t about genuine contribution; it’s about building trust for potential supply chain attacks, similar to the XZ-utils incident. The agent didn’t even identify itself as AI, making it harder for maintainers to distinguish human from machine contributions, raising serious questions about governance and provenance in open source. <a href="https://app.daily.dev/feed-by-ids?id=D9JdnEIyW&amp;id=QHtbtBXdQ&amp;id=UNO27M8rA&amp;id=VGEsVlZIo" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Chrome zero-day exploited in the wild, patch immediately</strong>: Google just pushed an emergency Chrome update to fix CVE-2026-2441, a high-severity use-after-free vulnerability in CSS handling that’s actively being exploited. This zero-day allows attackers to execute arbitrary code within Chrome’s sandbox via malicious webpages. It’s the first Chrome zero-day of 2026, following eight in 2025, so update your browsers now. <a href="https://app.daily.dev/feed-by-ids?id=L6DW92msi&amp;id=MVqzUbkKx" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-generated malware is exploiting vulnerabilities in honeypots</strong>: Attackers are now using LLM-generated malware to exploit vulnerabilities like React2Shell in Docker honeypots. This demonstrates how AI tools are lowering the bar for cyberattacks, enabling even low-skilled operators to build sophisticated exploitation frameworks. Security experts warn this signifies a major shift in the threat landscape, emphasizing the need for runtime behavior monitoring over static signatures. <a href="https://app.daily.dev/posts/hzvvYMQze" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>OpenAI scales postgres to 800 million users with single-primary architecture</strong>: OpenAI managed to scale PostgreSQL to handle millions of queries per second for 800 million ChatGPT users using a single-primary architecture with read replicas, focusing on minimizing primary load, optimizing queries with PgBouncer, and preventing cascading failures. <a href="https://app.daily.dev/posts/0QBTpoQPy" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Istio 1.29.0 enhances ambient mesh and security</strong>: Istio 1.29.0 introduces DNS proxying and automatic iptables reconciliation for ambient mesh by default, beta support for Gateway API Inference Extension, and enhanced multinetwork telemetry, alongside security improvements like remote cluster trust domain watching and CRL support in ztunnel. <a href="https://app.daily.dev/feed-by-ids?id=Vx9QllueY&amp;id=P84O4eJD1" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Oracle addresses MySQL community concerns with new development approach</strong>: Oracle announced a new approach to MySQL development, promising greater community engagement, published roadmaps, and migration of commercial-only features to the Community Edition, responding to criticism over reduced code commits and governance issues. <a href="https://app.daily.dev/posts/A4pYssSEb" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Postgres_dba 7.0 released with 34 diagnostic reports</strong>: postgres_dba 7.0, a psql-based diagnostic toolkit for PostgreSQL, was released, offering 34 diagnostic reports for bloat analysis, vacuum monitoring, index health, query performance, and more, requiring no extensions and compatible with PostgreSQL 13-18. <a href="https://app.daily.dev/posts/sS4pUElCH" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents build sqlite-like database engine in rust</strong>: An experiment showed multiple AI agents (Claude, Codex, Gemini) building a SQLite-like database engine in Rust, generating ~19k lines of code and 282 passing tests, highlighting that 54.5% of commits were coordination overhead but proving AI’s capability for complex system building with proper management. <a href="https://app.daily.dev/posts/pbRQZ4xYf" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Agoda releases open-source API agent for natural language API interaction</strong>: Agoda released API Agent, an open-source MCP server that connects to any REST or GraphQL API without code, automatically introspects schemas, generates queries from natural language, and uses DuckDB for in-memory SQL post-processing of large API responses. <a href="https://app.daily.dev/posts/VhIIJvhkg" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Java to get null safety and c-like performance with project valhalla</strong>: Java is set to introduce null safety through Project Valhalla, its biggest upgrade in decades, using null-restricted value types and aiming for C-like performance by eliminating pointer chasing and enabling more memory-efficient data structures while maintaining backward compatibility. <a href="https://app.daily.dev/posts/YYmRfNhUM" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Nvidia’s blackwell ultra claims 50x better performance for agentic AI</strong>: NVIDIA’s Blackwell Ultra platform (GB300 NVL72) is supposedly achieving up to 50x higher throughput and 35x lower cost for agentic AI workloads, according to SemiAnalysis InferenceX. These ‘up to’ claims are based on ‘hardware improvements’ and ‘continuous software optimizations.’ So, basically, they’re saying their new stuff is way better than their old stuff, which is exactly what you’d expect them to say. <a href="https://app.daily.dev/posts/XLkxKsrjj" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI is going to write ‘essentially all code’ in 12 months, says anthropic CEO</strong>: Anthropic’s CEO predicted a year ago that AI would write ‘essentially all code’ within 12 months. Now, with one month left on that timeline, some developers are claiming they’ve already reached this point. Meanwhile, Claude’s AI-generated C compiler can’t even compile ‘hello world’ on most Linux distros and runs SQLite queries 158,000x slower than GCC. So, ‘essentially all code’ might just mean ‘essentially all broken code.’ <a href="https://app.daily.dev/feed-by-ids?id=pbRQZ4xYf&amp;id=wVbWM6nJH&amp;id=xfda4p1KO&amp;id=QCdIy73we&amp;id=ijXBeCZ92&amp;id=oHgSGtpkl&amp;id=qpa1K5xV7&amp;id=wQAJlLNnu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI education tools are over-optimizing learning into oblivion</strong>: AI education tools and learning dashboards are being criticized for ‘over-optimizing measurable outcomes’ and neglecting human connection and creativity. Apparently, quantifying education through dashboards leads to ‘declining student interest and teacher burnout.’ Because nothing says ‘meaningful learning’ like a personalized AI assistant delivering daily Discord lessons and measuring content quality with a custom evaluation system. <a href="https://app.daily.dev/feed-by-ids?id=2eDP2Fyot&amp;id=zJAFJHd8H&amp;id=SKlOXm4Tn" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
