<h2>Must know</h2>
<ul>
<li><strong>Ladybird browser pivots to Rust with AI assistance</strong>: The Ladybird independent web browser project is adopting Rust as its C++ successor, primarily for memory safety, after a year-long detour with Swift. Lead developer Andreas Kling used AI agents like Claude Code and Codex to translate 25,000 lines of C++ to Rust in just two weeks, a task estimated to take months manually. This human-directed AI workflow resulted in byte-for-byte identical output and zero test regressions, demonstrating AI’s potential for large-scale, human-supervised code refactoring, as reported by Changelog News. <a href="https://app.daily.dev/feed-by-ids?id=bavH6r0ZC&amp;id=1QrlHS5ue" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-augmented FortiGate firewall compromises reported by AWS</strong>: AWS Threat Intelligence detected a financially motivated, Russian-speaking threat actor compromising over 600 FortiGate firewall instances across 55+ countries. The attackers, described as technically unsophisticated, leveraged generative AI tools for reconnaissance, exploitation instructions, scripting, and operational notes. This highlights how AI can enable less skilled actors to operate at scale, though AWS notes that basic security hygiene like MFA and restricted management interface exposure would have prevented most attacks. <a href="https://app.daily.dev/posts/azBzuZltn" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Uber open-sources uForwarder Kafka consumer proxy</strong>: Uber has open-sourced uForwarder, a push-based Kafka consumer proxy designed to handle trillions of messages daily across over 1,000 services. This system replaces direct Kafka client implementations with a gRPC-driven push interface, centralizing offset management and simplifying consumer logic. Key features include context-aware routing, an out-of-order commit tracker, and an adaptive auto-rebalancer, aiming to improve hardware utilization and reduce consumer lag. <a href="https://app.daily.dev/posts/NOSp6Xv2e" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>GitHub introduces Agent HQ for unified AI coding agents</strong>: GitHub announced Agent HQ, a unified platform for coding agents like Claude, Codex, and GitHub Copilot, allowing users to run and compare multiple frontier agents simultaneously with a single subscription. <a href="https://app.daily.dev/posts/CxN7d4Xat" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Kubernetes 1.35 release signals AI operating system role</strong>: Kubernetes v1.35 (‘Timbernetes’) introduces features like workload-aware scheduling for distributed training and stable in-place Pod resource resize, strengthening its role as an operational substrate for AI/ML workloads. <a href="https://app.daily.dev/feed-by-ids?id=mAyDHBOpw&amp;id=nytQTToPj" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI releases gpt-realtime-1.5 model</strong>: OpenAI released gpt-realtime-1.5 through its Realtime API, focusing on strengthening voice workflows with improved instruction following, tool calling, and multilingual accuracy for low-latency speech-to-speech interactions. <a href="https://app.daily.dev/posts/6I1BavGOr" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>ZeroDayRAT mobile spyware emerges, bypassing MFA</strong>: ZeroDayRAT, a new mobile spyware sold openly on Telegram, targets Android 5-16 and iOS up to 26, offering capabilities like keylogging, SMS control (for MFA bypass), and credential theft, commoditizing nation-state-level tools. <a href="https://app.daily.dev/posts/Q0yMWpjwU" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Firefox 148 release with AI controls</strong>: Firefox 148 introduces an ‘AI kill switch’ under Settings &gt; AI Controls, allowing users to disable all AI-powered features or selectively block individual ones, with Mozilla promising disabled features will not be re-enabled by future updates. <a href="https://app.daily.dev/posts/UNAL37VrS" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cohere releases Aya Tiny multilingual models</strong>: Cohere released the Aya Tiny model family, small (~3.3B parameter) multilingual language models designed to support over 70 languages, including low-resource ones, and run on mobile devices. <a href="https://app.daily.dev/posts/QusoaclzE" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Starkiller phishing kit bypasses MFA</strong>: Starkiller, a Phishing-as-a-Service (PhaaS) toolkit, bypasses MFA and traditional phishing detection by live-proxying real login pages through attacker-controlled cloud infrastructure, capturing credentials and authenticated session tokens. <a href="https://app.daily.dev/posts/9TGsm3aiu" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>OpenAI updates AI coding evaluation standard</strong>: OpenAI, in a move that absolutely no one saw coming, has decided its previous benchmark, SWE-bench Verified, is ‘saturated’ and ‘contaminated.’ So, naturally, they’re now recommending their <em>new</em> benchmark, SWE-bench Pro. Because when your own test shows your models are getting too good, you just invent a harder one, right? It’s not like they have a vested interest in constantly moving the goalposts to prove their AI is still ‘frontier.’ <a href="https://app.daily.dev/posts/oZGyDrjcX" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google DeepMind claims image generation breakthrough</strong>: A tweet from @rryssf_ claims Google DeepMind ‘just solved one of the dirtiest problems in image generation’ with an ‘embarrassingly elegant’ fix. Details? None. Official announcement? Crickets. So, we’re supposed to believe a vague, hype-filled social media post about a ‘breakthrough’ with zero technical specifics. Sounds like another day in AI land, where ‘breakthrough’ often means ‘we tweaked a parameter and it looks slightly better now.’ <a href="https://app.daily.dev/posts/fzJvtM3B4" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic accuses Chinese AI labs of industrial-scale model distillation</strong>: Anthropic is publicly accusing Chinese AI labs of ‘industrial-scale distillation attacks’ on its Claude models, claiming 24,000 fraudulent accounts and 16 million exchanges to extract capabilities. This, from the company facing multiple copyright lawsuits for training on scraped content. The irony is so thick you could distill it. And, of course, they’re framing it as a ‘national security threat’—because nothing says ‘we’re worried about competition’ like waving the flag. <a href="https://app.daily.dev/posts/YcZmCF704" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
