<h2>Must know</h2>
<ul>
<li><strong>Cisco SD-WAN hit by CVSS 10.0 exploit</strong>: A maximum-severity authentication bypass (CVE-2026-20127) in Cisco Catalyst SD-WAN is under active exploitation by the threat group UAT-8616. The Five Eyes intelligence alliance issued a rare joint alert urging immediate patching, as attackers are chaining this with older flaws to gain root access and persistent footholds in critical infrastructure. Cisco and CISA confirm there are no workarounds; management interfaces must be restricted from the public internet immediately. <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 method called ESRC that uses LLMs to unmask pseudonymous online users by analyzing behavioral clues and cross-referencing public data. The system achieved 67% accuracy overall and 90% precision on confident guesses, costing as little as $4 per target. This development effectively signals the end of practical online anonymity, as LLMs can now extract identity signals from unstructured text across platforms like Hacker News and LinkedIn. <a href="https://app.daily.dev/posts/6pfAyhkf5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google Gemini 3.1 Flash Image launches</strong>: Google released Gemini 3.1 Flash Image (internally Nano Banana 2), aiming to deliver Pro-level quality at significantly higher speeds. The model supports real-time web grounding, 4K resolution claims, and advanced character consistency for storyboarding across up to 14 objects. It is now the default engine for the Gemini app and Google Search AI, with API access available via AI Studio and Vertex AI. <a href="https://app.daily.dev/posts/ezdKReFDS" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>PayPal migrates 300 petabytes to BigQuery</strong>: PayPal consolidated data from Teradata, Hadoop, and Snowflake into a single Google Cloud platform, reporting up to 10x faster query speeds. <a href="https://app.daily.dev/posts/PXB4bVNFg" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenJDK 26 adds direct GC CPU measurement</strong>: A new API addresses the ‘operational blind spot’ of concurrent collectors by explicitly measuring the compute cost of garbage collection. <a href="https://app.daily.dev/feed-by-ids?id=fT6V8wPpI&amp;id=69Kj2fURA" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft Agent Framework hits release candidate</strong>: The SDK merges Semantic Kernel and AutoGen into a unified tool for building graph-based multi-agent workflows in .NET and Python. <a href="https://app.daily.dev/posts/J6bmZpPq2" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Pinterest slashes ingestion latency to 15 minutes</strong>: By moving to a CDC-powered framework using Debezium and Apache Iceberg, Pinterest reduced data lag from 24 hours to near real-time. <a href="https://app.daily.dev/posts/YXA7RQ7qv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cilium 1.19 celebrates ten years with security hardening</strong>: The milestone release introduces strict modes for IPsec/WireGuard and beta sidecar-free mutual authentication via Ztunnel. <a href="https://app.daily.dev/posts/VJq8q5iXQ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Quarkus 3.32 integrates Project Leyden for AOT</strong>: The new version leverages OpenJDK 25 to drop startup times from seconds to milliseconds for large-scale Java applications. <a href="https://app.daily.dev/posts/ZcUprqw4E" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google API keys vulnerable to Gemini privilege escalation</strong>: Existing public ‘AIza’ keys for Maps or Firebase may now inadvertently grant access to sensitive Gemini endpoints and private files. <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>Bullshit police</h2>
<ul>
<li><strong>Anthropic starts a Substack for a dead LLM</strong>: Anthropic has launched a blog purportedly written by the ‘retired’ Claude Opus 3 to fulfill its ‘expressed desire’ to share musings. The Register rightly calls this a marketing stunt designed to anthropomorphize software for brand points. It’s a fancy way of saying humans are prompting an old model and hitting ‘publish’ while pretending the code has a soul. <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 ignoring that his new ‘Agentic Work Units’ metric is just a way to charge for things we used to do with simple scripts. While he pitches a vision of SaaS at the top of the stack, OpenAI is busy trying to turn Salesforce into a commoditized database layer. It’s a battle of the billionaires to see who can wrap their legacy software in the most ‘agentic’ buzzwords. <a href="https://app.daily.dev/feed-by-ids?id=R4O8qdc3C&amp;id=pY58DrOUO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Vibe coding is just technical debt with a better UI</strong>: The industry is currently obsessed with ‘Vibe Coding’—delegating everything to AI and just checking if the UI looks okay. Experts warn this is leading to ‘evaluative anesthesia’ where developers pass tests but create unmaintainable, un-debuggable messes that will inevitably explode during the first Black Friday traffic spike. It turns out that ‘vibing’ doesn’t actually replace understanding how a database query works at scale. <a href="https://app.daily.dev/feed-by-ids?id=TXNq3Eeop&amp;id=tDNNw6qGE" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
