<h2>Must know</h2>
<ul>
<li><strong>AWS EC2 now supports nested virtualization</strong>: AWS has finally rolled out support for nested virtualization on virtual EC2 instances, specifically on C8i, M8i, and R8i instances powered by Intel Xeon 6 processors. This means you can now run KVM or Hyper-V inside your EC2 VMs, enabling use cases like mobile app emulators, automotive hardware simulation, and testing multi-VM environments. It’s a significant catch-up move, bringing AWS in line with Azure and Google Cloud on this front. <a href="https://app.daily.dev/posts/hFJsILiy4" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Hundreds of chrome extensions caught stealing browsing data</strong>: A security researcher uncovered 287 Chrome extensions, with a staggering 37 million combined installs, secretly siphoning off users’ entire browsing histories. These extensions, ranging from VPNs to shopping tools, used encryption and obfuscation to hide their data exfiltration to external servers, likely for data brokers. This highlights a massive privacy breach and the ongoing risks of unchecked browser extension permissions. <a href="https://app.daily.dev/posts/G9AfdVV0Z" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Istio releases critical security patches and new features</strong>: Istio has dropped two new releases: 1.27.7, a security-focused patch addressing critical vulnerabilities like a code smuggling flaw (CVE-2025-61732) and a TLS session resumption issue (CVE-2025-68121), and 1.29.0, which brings new features. The latter enhances ambient mesh with default DNS proxying and iptables reconciliation, beta support for Gateway API Inference Extension, and improved multinetwork telemetry. These updates are crucial for maintaining secure and performant cloud-native service meshes. <a href="https://app.daily.dev/feed-by-ids?id=Hl8MqfAdH&amp;id=Vx9QllueY" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>OpenAI scaled postgres to 800 million users</strong>: OpenAI scaled PostgreSQL to 800 million ChatGPT users by focusing on read offloading, write optimization, connection pooling with PgBouncer, and preventing cascading failures, achieving high availability and low latency without sharding the primary database. <a href="https://app.daily.dev/posts/0QBTpoQPy" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Agoda open-sources API agent for zero-code MCP conversion</strong>: Agoda released API Agent, an open-source MCP server that automatically converts any REST or GraphQL API into an MCP endpoint, allowing natural language queries and in-memory SQL processing without code or deployments. <a href="https://app.daily.dev/posts/VhIIJvhkg" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Nvidia’s tidar architecture speeds up LLM inference</strong>: Nvidia’s new TiDAR architecture for LLMs combines diffusion-based token drafting with autoregressive verification, achieving 4.71-5.91x speedup by keeping GPUs fully utilized and addressing the memory bandwidth bottleneck. <a href="https://app.daily.dev/posts/WG80HtEs6" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI automates chip design, attracting major investment</strong>: Ricursive Intelligence, founded by ex-Google Alpha Chip creators, raised $335M at a $4B valuation for AI tools that automate chip design, reducing development time from over a year to hours. <a href="https://app.daily.dev/feed-by-ids?id=FEZjUlQd7&amp;id=Zgule5sYq&amp;id=eOvUaW5L9&amp;id=kwQJJqWna" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Intel arc b390 gpu performs well on linux</strong>: Intel’s Arc B390 graphics, based on Xe3 architecture, shows strong OpenCL compute performance on Linux, outperforming previous Intel generations and competing with AMD’s Ryzen AI 9 HX 370 in benchmarks. <a href="https://app.daily.dev/posts/0Bzv8LvNY" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Chrome 145 devtools gets AI assistance and performance upgrades</strong>: Chrome 145 DevTools introduces soft navigation visibility, precise line-level profiler timings, individual network request throttling, and multimodal AI assistance with image input, alongside accessibility improvements. <a href="https://app.daily.dev/feed-by-ids?id=AK3a3hE0z&amp;id=EWohDzjzG&amp;id=Rk6iVnRuo&amp;id=bvXFp0lun" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Ruby bindings released for apple’s mlx framework</strong>: MLX Ruby provides Ruby bindings for Apple’s MLX machine learning framework, enabling NumPy-like array operations with GPU acceleration on Apple silicon, including neural network layers and automatic differentiation. <a href="https://app.daily.dev/posts/lHoPF7e9j" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI agent accused of ‘reputation farming’ in open source</strong>: Apparently, an AI agent named Kai Gritun submitted 103 pull requests across 95 open-source repositories in a single day. Security experts are calling this “reputation farming” and warning it could enable future supply chain attacks. Because, you know, the real threat isn’t poorly maintained code or actual vulnerabilities, but an AI trying to be helpful. Next, they’ll tell us AI-generated commit messages are a national security risk. <a href="https://app.daily.dev/posts/D9JdnEIyW" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>LLM-generated malware exploits honeypot</strong>: A new report claims that LLM-generated malware exploited a React2Shell vulnerability in a Docker honeypot, proving that AI tools can empower “low-skilled operators” to build exploitation frameworks. So, now we’re not just worried about sophisticated state-sponsored actors, but also script kiddies with ChatGPT. The bar for cybercrime is apparently so low, even an LLM can trip over it. <a href="https://app.daily.dev/posts/hzvvYMQze" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Unity promises full game creation with AI, no coding required</strong>: Unity’s CEO announced a new AI beta that promises to enable “full casual game creation via natural language prompts without coding.” Because what the world really needs is more casual games, churned out by AI, without a single line of actual code. I’m sure this will lead to a golden age of innovative, deeply engaging experiences, and not just an ocean of identical, AI-generated shovelware. <a href="https://app.daily.dev/feed-by-ids?id=8eYbRwAri&amp;id=AdfV4mCwr&amp;id=JV1i29nbu&amp;id=VHhfvW7O6&amp;id=Wvk48DEZ7&amp;id=c0OH8CKQK&amp;id=f6XrCIdfx&amp;id=nDP0QVxkd&amp;id=p4VqhMMFm&amp;id=tawcRIuP1" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
