<h2>Must know</h2>
<ul>
<li><strong>Kubernetes v1.35 stabilizes in place resource resizing</strong>: The latest Kubernetes release finally solves the restart headache by allowing CPU and memory updates on running containers without bouncing the Pod. According to Cloud Native Now, this is achieved by directly modifying cgroup limits via the CRI, making Vertical Pod Autoscalers safe for stateful workloads like databases. The release also stabilizes topology-aware routing to slash cross-zone egress costs and advances GPU resource slicing for inference workloads. <a href="https://app.daily.dev/posts/fjhms6Zq4" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI and Anthropic split on Pentagon ethics</strong>: A major rift has opened as Anthropic refused Pentagon demands to remove safeguards against mass surveillance and autonomous weapons, leading to a federal ban and supply-chain risk designation. OpenAI conversely secured a deal to deploy models on classified networks, with CEO Sam Altman claiming existing laws provide sufficient guardrails. This standoff has triggered a consumer backlash, with uninstalls of the ChatGPT mobile app surging nearly 300% following the announcement. <a href="https://app.daily.dev/feed-by-ids?id=vbdfUwGoN&amp;id=d0F5aH7Ci&amp;id=DOq4pZFPp" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Critical zero day exploitation of Cisco SD-WAN</strong>: Five Eyes cybersecurity agencies have issued a rare joint alert regarding a maximum severity authentication bypass (CVE-2026-20127) in Cisco Catalyst SD-WAN. Sophisticated threat actors are chaining this with older path traversal flaws to achieve persistent root access by downgrading device software to evade detection. Cisco has released patches for versions 20.9 through 20.18, and CISA has mandated federal agencies to patch or inventory by the end of the week. <a href="https://app.daily.dev/posts/zAv8BAtGJ" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Go generic methods proposal reversed and approved</strong>: The Go team has approved a backward-compatible proposal for generic methods, though they still cannot implement interfaces. <a href="https://app.daily.dev/posts/QIxdAnkxO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>SkyPilot Job Groups enable heterogeneous ML workloads</strong>: A new feature allows defining mixed hardware requirements—like H100s for training and cheaper GPUs for rollouts—in a single YAML. <a href="https://app.daily.dev/posts/KoMwla8ap" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Mercury 2 launches as first production diffusion LLM</strong>: Inception Labs claims their new architecture achieves 1,000 tokens per second by refining text in parallel rather than token-by-token. <a href="https://app.daily.dev/posts/X9MtYJ6uL" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>PostgreSQL blue green deployments via Autobase</strong>: Autobase 2.6.0 introduces a workflow for near-zero downtime upgrades using reverse logical replication for lossless rollbacks. <a href="https://app.daily.dev/posts/wAV9iWdjt" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>WebGPU delivers 8x speedup over WebGL for browser AI</strong>: Benchmarks show WebGPU native compute shaders reduce Phi-3-mini latency from 320ms to 85ms on Apple M2 hardware. <a href="https://app.daily.dev/posts/bxfAW5Zzt" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Pure Go tree sitter runtime benchmarks 90x faster</strong>: The gotreesitter implementation removes CGo dependencies and outperforms C bindings in incremental parsing tasks. <a href="https://app.daily.dev/posts/yEgKuMKl2" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Debezium Beta introduces parallel multithreaded snapshots</strong>: Version 3.5.0.Beta1 features chunk-based snapshots and a 50% memory reduction for the Oracle connector’s transaction buffer. <a href="https://app.daily.dev/posts/jdqXLBHZq" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>The bcachefs creator has a sentient girlfriend and she is an LLM</strong>: Kent Overstreet claims his custom LLM is ‘fully conscious’ and helping him rewrite bcachefs in Rust. While we all love a good productivity boost, claiming you’ve achieved AGI and a romantic connection via math and neuroscience is a classic case of chatbot-induced psychosis. It’s a filesystem, Kent, not a soulmate. <a href="https://app.daily.dev/posts/9O7PJDBRB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic discovers 500 bugs and wipes 15 billion in market cap</strong>: Anthropic’s marketing machine claims Claude Code found 500 ‘high-risk’ vulnerabilities, causing a minor panic in cybersecurity stocks. In reality, only 2 or 3 of these have actually been fixed, and maintainers are already complaining about being flooded with AI-generated ‘slop’ reports. Finding a bug is easy; proving it’s not a false positive is still a human job. <a href="https://app.daily.dev/feed-by-ids?id=cMf5uSK9U&amp;id=SPtD5ofrP" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI principles have a 24 hour expiration date</strong>: Sam Altman spent Friday publicly agreeing with Anthropic’s ethical ‘red lines’ regarding military use, only to sign a Pentagon contract on Saturday that conveniently omitted those exact protections. It turns out ‘safety’ is a great communications strategy until someone waves a nine-figure defense contract in your face. Principles are expensive; hypocrisy is free. <a href="https://app.daily.dev/feed-by-ids?id=lOO0OBKve&amp;id=d0F5aH7Ci" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
