<h2>Must know</h2>
<ul>
<li><strong>AWS doubles down on data centers for AI</strong>: AWS reported its strongest quarterly growth in over three years, driven by traditional cloud migration and AI workloads. The company plans to invest $200 billion in 2026 to double its data center capacity by the end of 2027, primarily for AI infrastructure. This massive investment signals AWS’s commitment to maintaining its lead in the cloud and AI compute space, as reported by AWS itself. <a href="https://app.daily.dev/posts/1BAlXIloL" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft open-sources Rust-based sandboxing project LiteBox</strong>: Microsoft has open-sourced LiteBox, an experimental library OS written in Rust, designed for application sandboxing with a minimal attack surface. It aims to provide stronger isolation than containers without the overhead of full VMs, supporting confidential computing and running unmodified Linux applications. This is a significant development for platform and infrastructure engineers focused on security and efficient isolation, directly from Microsoft. <a href="https://app.daily.dev/feed-by-ids?id=jIlb9nmzq&amp;id=0PdnWmIzP" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Supply chain attacks target dYdX protocol and kernel drivers</strong>: Socket’s Threat Research Team uncovered a supply chain attack on dYdX protocol packages on npm and PyPI, exfiltrating crypto wallet credentials and device fingerprints. Separately, a security researcher found critical vulnerabilities in Tower of Fantasy’s anti-cheat kernel driver, allowing arbitrary process termination and EDR evasion. These incidents highlight persistent threats to software supply chains and the exploitation of low-level system components. <a href="https://app.daily.dev/feed-by-ids?id=bwRxiW1f3&amp;id=iRYZ9eAV6" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Go-based Bifrost outperforms Python’s LiteLLM for LLM gateways</strong>: For production LLM gateways requiring high concurrency, Go-based Bifrost maintains P99 latency around 1.6-1.7 seconds under sustained load, while Python’s LiteLLM degrades significantly, demonstrating Go’s advantage for performance-critical infrastructure. <a href="https://app.daily.dev/posts/hETwaFbwz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Snowflake integrates pg_lake for open data standards</strong>: Snowflake integrated <code>pg_lake</code>, an open-source PostgreSQL extension, into its Snowflake Postgres service, enabling direct SQL interaction with data lakehouses and Apache Iceberg tables, emphasizing interoperability and open standards. <a href="https://app.daily.dev/posts/kBWxo6cQz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NVIDIA NVFP4 accelerates AI performance</strong>: NVIDIA introduced NVFP4, a 4-bit floating-point precision format for AI training and inference, achieving up to 3x faster performance than FP8 on Blackwell Ultra GPUs while maintaining accuracy. <a href="https://app.daily.dev/posts/2zfw3Y2wB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux kernel scheduling enhancements proposed</strong>: Cambridge researchers developed Latency-Aware Group Scheduling (LAGS) to boost Kubernetes cluster capacity by 10-20%, while a new patch series, DHEI, aims for dynamic CPU resource partitioning without reboots, both optimizing Linux kernel resource management. <a href="https://app.daily.dev/feed-by-ids?id=cc5mbXUoZ&amp;id=sIpmbSkbv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>KubeCon + CloudNativeCon Europe 2026 to feature extensive OpenTelemetry content</strong>: KubeCon + CloudNativeCon Europe 2026 in Amsterdam will host 21 main conference talks and 24 Observability Day sessions dedicated to OpenTelemetry, covering topics from Collector updates to cost optimization strategies. <a href="https://app.daily.dev/posts/SZ5x4LtZK" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cisco introduces deep network model for AI-driven network operations</strong>: Cisco launched its Deep Network Model, an AI system trained on 100 million tokens of networking knowledge, outperforming general-purpose LLMs by 20% on networking benchmarks for troubleshooting and operational tasks. <a href="https://app.daily.dev/posts/mxQWg838o" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Zed’s community sprint ships 66 Git improvements</strong>: Zed’s ‘Let’s Git Together’ community sprint resulted in 66 improvements to Git functionality, including major features like file history view and remote management UI, demonstrating effective open-source collaboration. <a href="https://app.daily.dev/posts/o4WTqP3cO" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI identifies 500 high-severity vulnerabilities, human verification pending</strong>: Anthropic’s Claude Opus 4.6 supposedly found 500 high-severity vulnerabilities in open-source software, operating autonomously. Because nothing says ‘trustworthy’ like an AI finding bugs without human oversight, and then needing humans to verify its findings before reporting them. What could possibly go wrong? <a href="https://app.daily.dev/posts/DDZLtuMLV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI code generation tools blamed for open-source project layoffs</strong>: Tailwind CSS, despite peak popularity, laid off 75% of its engineering team, blaming AI tools for generating code without developers visiting their documentation site. Apparently, AI is so good it’s making documentation irrelevant, thus collapsing business models reliant on traffic. Who knew AI’s biggest threat was to ad revenue, not jobs? <a href="https://app.daily.dev/posts/uR7Q2jMhV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Postman transforms into an ‘everything app’</strong>: Postman’s latest updates turn it into an ‘everything app’ with AI agent builders, managed cloud servers, and internal terminals. Because what developers really need is more feature creep, vendor lock-in, and performance issues from a tool that used to just test APIs. Curl and jq are still free, just saying. <a href="https://app.daily.dev/posts/O8Zfhee9v" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
