<h2>Must know</h2>
<ul>
<li><strong>Interlock ransomware hits Cisco firewalls via zero day</strong>: The Interlock ransomware group exploited a critical zero-day vulnerability (CVE-2026-20131) in Cisco Secure Firewall Management Center for 38 days before a patch was released. Amazon’s MadPot honeypot network uncovered the full attack chain, which allowed remote code execution via deserialization and carries a maximum CVSS score of 10. Security admins are urged by Cisco and Amazon researchers to patch immediately to prevent further exploitation by this RaaS group. <a href="https://app.daily.dev/posts/kS1FWAP6y" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>FBI seizes Handala sites after massive Stryker wipe</strong>: The FBI and DOJ have seized the websites of the pro-Iranian group Handala following a devastating cyberattack on medical tech giant Stryker. Handala reportedly compromised an admin account to hijack Microsoft Intune dashboards, remotely wiping tens of thousands of employee devices and knocking Stryker’s supply systems offline. CISA has issued an urgent alert recommending dual-administrator approval for high-impact Intune actions like device wipes to prevent similar supply chain paralysis. <a href="https://app.daily.dev/posts/N8kFe3mft" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cursor launches Composer 2 with self summarization</strong>: Cursor has released Composer 2, an in-house coding model that claims to outperform Claude Opus 4.6 on terminal-based benchmarks while being significantly cheaper. The model utilizes a ‘self-summarization’ reinforcement learning technique that compresses its context window to 1,000 tokens to maintain performance when limits are reached. Despite the impressive benchmarks reported by Cursor, internal discrepancies in scoring suggest these company-led claims should be verified by independent testing. <a href="https://app.daily.dev/posts/mIKxoNmwG" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Google AI Studio adds vibe coding for full stack app generation</strong>: A new Build Mode automates frontend, backend, and Cloud Run deployment from natural language prompts. <a href="https://app.daily.dev/posts/vDejXlEPp" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Heroku CLI v11 migrates to ESM and doubles slug limits</strong>: The architectural overhaul moves to Node 22 and increases slug sizes to 1GB to better support heavy AI/ML frameworks. <a href="https://app.daily.dev/feed-by-ids?id=rnRadhI60&amp;id=vu5ItVeI2&amp;id=2swWgPS9c" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Checkmarx One deploys AI agents for automated triage</strong>: The platform update uses five orchestrated agents to handle vulnerability discovery and code remediation at ‘machine speed.’ <a href="https://app.daily.dev/posts/PE8tBh7Qo" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Neo4j Aura Agent simplifies GraphRAG deployment</strong>: The new low-code platform allows developers to build AI agents using knowledge graphs and Text2Cypher without deep graph expertise. <a href="https://app.daily.dev/posts/mc7JPTHNY" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Proxmox 9 performance tip use VMs for stability and LXC for labs</strong>: New benchmarks show LXC uses 75% less RAM than VMs but risks host-wide kernel panics on shared Proxmox kernels. <a href="https://app.daily.dev/posts/3AOBWZq6l" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AWS SDK for Ruby adds bulk S3 directory transfers</strong>: The v1.215+ update introduces native methods for recursive directory uploads and downloads with built-in parallelism. <a href="https://app.daily.dev/posts/tlRI3NVwm" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>PHP 8.5 arrives with native URI extension and pipe operator</strong>: The ecosystem update also sees FrankenPHP outperforming PHP-FPM by nearly 4x in recent benchmarks. <a href="https://app.daily.dev/posts/9QGNRKN0e" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Meta claims AI is the hero of content moderation</strong>: Meta is bragging that its new AI tools detected 5,000 phishing attempts, conveniently ignoring that The Register pointed out these ‘innovations’—like flagging impossible login locations—have been standard enterprise security features for a decade. It’s a classic pivot: after burning $80 billion on a metaverse nobody wanted, they’re rebranding basic heuristic checks as ‘superintelligence’ to distract shareholders. Apparently, human moderators were just forbidden from using basic logic until the AI marketing budget kicked in. <a href="https://app.daily.dev/posts/Y6ptsDngz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>LangSmith rebrands Agent Builder to Fleet for enterprise vibes</strong>: LangSmith has decided that ‘Agent Builder’ wasn’t corporate enough, so they’ve rebranded it to ‘Fleet’ to make you feel like a high-powered admiral of a bunch of LLM scripts. They’re pitching ‘Agent Inboxes’ and ‘Claws’ credentials as the solution to scaling, but it’s mostly just adding a layer of Slack-bot management to the existing chaos of AI agents. It’s a lot of enterprise-grade paint on the same underlying problem of trying to keep non-deterministic bots from hallucinating in production. <a href="https://app.daily.dev/posts/fnry81udE" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The high reasoning AI that solved a six month bug</strong>: The tech world is swooning over a story where OpenAI’s Codex solved a screen flicker bug by ‘reading the source code,’ as if that isn’t exactly what we’ve been told these models do for years. The PrimeTime frames this as a ‘we’re so back’ moment for engineering, but it’s really just a reminder that we’re now celebrating when the expensive black box actually does its job instead of hallucinating. We’ve reached a point where an AI performing basic documentation lookup is considered a breakthrough worth a blog post. <a href="https://app.daily.dev/posts/qudWCnXkT" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
