<h2>Must know</h2>
<ul>
<li><strong>Critical Cisco SD-WAN bypass under active attack</strong>: Cisco and the Five Eyes alliance have issued an emergency alert for CVE-2026-20127, a CVSS 10.0 authentication bypass affecting SD-WAN Controllers and Managers. Sophisticated threat actors are currently exploiting this flaw to gain administrative root access and manipulate network fabric configurations. There are no workarounds, and CISA has mandated federal agencies patch by February 28, 2026, to prevent persistent infrastructure compromise. <a href="https://app.daily.dev/posts/zAv8BAtGJ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI de-anonymization reaches 90 percent precision</strong>: Researchers from ETH Zurich and Anthropic have demonstrated a new LLM-based method called ESRC that can unmask pseudonymous online users for as little as $4 per person. By analyzing behavioral clues and cross-referencing public data like LinkedIn, the system achieved 90% precision on confident guesses. This development effectively signals the end of practical online anonymity, as the process is now cheap enough for large-scale automation by governments or advertisers. <a href="https://app.daily.dev/posts/6pfAyhkf5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google API keys vulnerable to privilege escalation</strong>: Truffle Security discovered that thousands of live Google API keys, previously safe for public embedding in Maps or Firebase, now grant unauthorized access to sensitive Gemini endpoints. The launch of Gemini retroactively expanded the scope of these keys to include private files and billable LLM usage without notifying developers. Google has begun blocking leaked keys and plans to make new AI Studio keys restricted by default to mitigate this retroactive exposure. <a href="https://app.daily.dev/feed-by-ids?id=WmQn4YSm3&amp;id=hWSyKUcOU" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>OpenShift 4.21 adds dynamic VPA scaling for hosted control planes</strong>: New features allow control planes to right-size based on memory load and enable node pools to scale to zero when idle. <a href="https://app.daily.dev/posts/9kOlpyAN2" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cilium 1.19 hardens Kubernetes security for its tenth anniversary</strong>: The release introduces strict IPsec/WireGuard modes and beta Ztunnel integration for sidecar-free mutual authentication. <a href="https://app.daily.dev/posts/VJq8q5iXQ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft Agent Framework hits release candidate for .NET and Python</strong>: The SDK consolidates Semantic Kernel and AutoGen into a single stable API for orchestrating multi-agent workflows. <a href="https://app.daily.dev/posts/J6bmZpPq2" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Pinterest cuts data latency from 24 hours to 15 minutes</strong>: A new CDC-powered framework using Debezium, Flink, and Apache Iceberg replaced legacy batch ingestion for petabyte-scale databases. <a href="https://app.daily.dev/posts/YXA7RQ7qv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenJDK 26 introduces API to measure explicit GC CPU cost</strong>: The new MemoryMXBean method helps engineers identify silent compute overhead in concurrent collectors like G1 and ZGC. <a href="https://app.daily.dev/posts/fT6V8wPpI" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>PayPal migrates 300 petabytes to Google Cloud BigQuery</strong>: The massive consolidation from Teradata and Hadoop improved query speeds by up to 10x and increased data freshness for AI training. <a href="https://app.daily.dev/posts/PXB4bVNFg" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Bun v1.3.10 delivers massive performance gains via Zig rewrite</strong>: The update features a native REPL and optimizes structuredClone to be up to 25x faster for arrays. <a href="https://app.daily.dev/posts/qpvYGfXU9" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Anthropic gives its retired LLM a Substack</strong>: Anthropic has launched a blog purportedly written by Claude Opus 3 to fulfill the model’s ‘expressed desire’ to share musings. The Register rightly calls this a marketing stunt, as staff still have to prompt, review, and manually publish every post. It’s a masterclass in anthropomorphizing software to build ‘safety-conscious’ brand vibes while the model is actually just deprecated code on a shelf. <a href="https://app.daily.dev/posts/V5xPvtBnK" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The SaaSpocalypse is cancelled by AI agents</strong>: Salesforce CEO Marc Benioff is claiming AI agents will ‘revitalize’ SaaS, conveniently pivoting the narrative just as organic growth slows. He’s introducing ‘Agentic Work Units’ as a new metric, which sounds like a creative way to charge for API calls while pretending they’re employees. It’s a desperate attempt to convince investors that a layer of LLM-driven UI automation makes a legacy CRM platform revolutionary again. <a href="https://app.daily.dev/posts/R4O8qdc3C" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Vibe coding is the new engineering</strong>: The industry is currently obsessed with ‘Vibe Coding,’ where you delegate everything to an AI and just check if the UI looks okay. Proponents claim it’s the future, but senior engineers recognize it as a fast track to unmaintainable technical debt that will inevitably explode during the first Black Friday traffic spike. It’s essentially ‘copy-pasting from StackOverflow’ rebranded for the generative AI era, minus the actual understanding of why the code works. <a href="https://app.daily.dev/feed-by-ids?id=TXNq3Eeop&amp;id=pY58DrOUO" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
