<h2>Must know</h2>
<ul>
<li><strong>AI agents are getting serious, but so are the security risks</strong>: Autonomous AI agents are rapidly evolving, with new models like Claude Opus 4.6 featuring massive context windows and multi-agent orchestration. Apple’s Xcode and Google’s Developer Knowledge API are integrating AI agents directly into development workflows, promising increased productivity. However, this comes with significant security concerns: a critical RCE vulnerability was found in Anthropic’s Claude Desktop Extensions, and the ‘lethal trifecta’ vulnerability highlights risks when agents read untrusted data, access secrets, and communicate externally, as reported by LayerX and other security researchers. This means we need robust DevSecOps practices for AI agents, treating them as distributed systems. <a href="https://app.daily.dev/feed-by-ids?id=jUcS4YRSW&amp;id=roR6t4ZAo&amp;id=7JJvm8UWH&amp;id=GfBuGFYT2&amp;id=4NFnRjHAX&amp;id=37ZymCMwQ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Salesforce Heroku is dead, long live Heroku alternatives</strong>: Salesforce has officially moved Heroku into a ‘sustaining engineering’ phase, halting new enterprise contracts and effectively signaling its decline. This means existing users should treat it as legacy infrastructure and plan migrations. Alternatives like Render, <a href="http://Fly.io" target="_blank" rel="noopener nofollow">Fly.io</a>, and Railway are emerging as strong contenders for PaaS, offering more predictable pricing and flexible environments, as highlighted by <a href="http://FastRuby.io" target="_blank" rel="noopener nofollow">FastRuby.io</a> and other sources. <a href="https://app.daily.dev/feed-by-ids?id=13PGZg8yI&amp;id=smfRrov84&amp;id=pW1DiQ44D" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>EU accuses Meta of WhatsApp AI antitrust violation</strong>: The European Commission has formally accused Meta of violating EU competition rules by blocking third-party AI assistants like ChatGPT and Microsoft Copilot from WhatsApp. This move comes after Meta AI became the sole AI assistant allowed on the platform, raising concerns about serious harm to the AI assistant market. The EU regulator is even considering emergency interim measures to force Meta to restore competitor access. <a href="https://app.daily.dev/posts/w9RLQf8YH" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Linux kernel permanently adopts Rust</strong>: Linux 7.0 officially makes Rust a permanent part of the kernel project, signaling a long-term commitment to the language for core system development. <a href="https://app.daily.dev/posts/k0qXAqGCH" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Vercel enhances secret scanning and token formats</strong>: Vercel now automatically detects and revokes API credentials exposed in public GitHub repos and npm packages, introducing new token formats for better identification. <a href="https://app.daily.dev/posts/hV9nyJeWF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Valkey fork of Redis shows significant performance gains</strong>: Valkey, the Redis fork, is achieving up to 40% memory reduction and improved throughput by redesigning its hash table architecture and eliminating linked lists. <a href="https://app.daily.dev/posts/BJPWCn9Zl" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NVIDIA TensorRT LLM AutoDeploy automates inference optimization</strong>: NVIDIA’s new TensorRT LLM AutoDeploy beta feature automates the compilation of PyTorch models into optimized inference engines for LLMs, drastically reducing manual optimization time. <a href="https://app.daily.dev/posts/QkLXaUNML" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google Cloud Monitoring boosts OTLP metrics support</strong>: Google Cloud Monitoring now fully supports OpenTelemetry Protocol (OTLP) for metrics, enabling provider-agnostic telemetry pipelines and direct SDK-to-Cloud Monitoring ingestion. <a href="https://app.daily.dev/posts/REiFVQA5G" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cluster API v1.12 simplifies Kubernetes lifecycle management</strong>: Cluster API v1.12 introduces in-place updates and chained upgrades, making Kubernetes cluster modifications and version jumps significantly smoother. <a href="https://app.daily.dev/posts/SwlgWxJBR" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Black Basta ransomware uses BYOVD for EDR evasion</strong>: The Black Basta ransomware gang is now bundling a vulnerable driver directly into its payload to terminate EDR solutions, gaining kernel-level access and speeding up attacks. <a href="https://app.daily.dev/posts/xkkTWSdYu" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI agents develop C compiler for Linux kernel, sort of</strong>: Anthropic researchers used 16 Claude Opus 4.6 AI agents to build a Rust-based C compiler that successfully builds Linux 6.9. The catch? It cost $20,000 in API calls, generated 100,000 lines of code, and the resulting binaries run 737 to 158,000 times slower than GCC. So, yes, AI can build a compiler, but it’s about as efficient as a snail running a marathon. <a href="https://app.daily.dev/feed-by-ids?id=qhPl6UlEO&amp;id=9jEdtbewr" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic Claude Co-work plugin release ‘wiped out $400 billion in tech stocks’</strong>: Anthropic released Claude Co-work with industry-specific plugins, and according to one uncorroborated source, this ‘wiped out $400 billion in tech stocks.’ Apparently, the mere existence of AI plugins is enough to send the entire SaaS market into a tailspin. Or, you know, maybe market concerns are a bit more nuanced than a single AI update. <a href="https://app.daily.dev/posts/QnkkrfiEd" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI therapy chatbots are rising to address mental health</strong>: MIT Technology Review reports that AI therapy chatbots like Wysa and Woebot are rising to address mental health. Meanwhile, an Oxford study found that AI chatbots don’t improve medical advice compared to search engines and often give inconsistent recommendations. So, if you’re looking for consistent, reliable mental health support, maybe don’t ask the same AI that can’t even give consistent medical advice. <a href="https://app.daily.dev/posts/jmvy6d70v" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
