<h2>Must know</h2>
<ul>
<li><strong>Google expands AI editing and search with new models</strong>: Google is rolling out AI-driven, prompt-based photo editing in Google Photos to India, Australia, and Japan, using the Nano Banana AI model for on-device processing. This feature allows users to edit images with natural language commands and includes C2PA Content Credentials for transparency. Concurrently, Google is integrating AI Mode conversations into AI Overviews in Search, making Gemini 3 the default AI model globally for interactive search experiences. These moves, reported by Google’s official communications, aim to enhance accessibility and conversational capabilities across its AI offerings. <a href="https://app.daily.dev/feed-by-ids?id=QDh2xpUc9&amp;id=rbuaaR7gi" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents are transforming enterprise workflows</strong>: AI agents are moving beyond chatbots to become embedded ‘coworkers’ within enterprise systems, shifting from cloud data centers to local environments. This transformation, highlighted by industry analysis, introduces significant governance challenges, including ‘agent sprawl’ and the need for clear accountability. Organizations deploying AI governance frameworks achieve 12 times more projects in production, emphasizing the critical role of secure deployment and robust oversight. The shift is driving new ‘Lakebase’ architectures, with 80% of new serverless databases now created by AI agents, according to a trade report. <a href="https://app.daily.dev/feed-by-ids?id=yBxtGCVqr&amp;id=6kYqgkgUF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NVIDIA open-sources FastGen for diffusion model acceleration</strong>: NVIDIA has released FastGen, an open-source library designed to accelerate diffusion models by 10x to 100x without sacrificing quality. This plug-and-play solution, detailed on NVIDIA Developer, uses distillation techniques and supports models up to 14 billion parameters, enabling real-time video generation and interactive world modeling. FastGen also includes reproducible benchmarks and optimizations like FSDP2 and AMP, aiming to lower hardware and time barriers for deploying large diffusion models across various domains, including weather modeling. <a href="https://app.daily.dev/posts/QgRRffHqE" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Ingress-nginx project is being archived, migrate to Cilium</strong>: The ingress-nginx project will be archived in early 2026, ending active maintenance and security updates, with users advised to migrate to Cilium Ingress for a drop-in replacement or Cilium’s Gateway API for advanced features and eBPF-powered performance. <a href="https://app.daily.dev/posts/PptBbhnAu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Rust Foundation outlines 2026-2028 strategic plan</strong>: The Rust Foundation released its 2025 Annual Report and 2026-2028 strategic plan, prioritizing stable infrastructure, maintainer support, responsible adoption, and community engagement, having raised $5.1 million in 2025 with $2.7 million invested directly into the Rust Project. <a href="https://app.daily.dev/posts/DRpplNBlF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>WhatsApp adopts Rust for media security overhaul</strong>: WhatsApp replaced 160,000 lines of C++ code with 90,000 lines of Rust in its media handling stack, deploying the Kaleidoscope library globally across Android, iOS, web, and wearables to validate media files and detect malicious patterns, marking the largest Rust deployment for a consumer app. <a href="https://app.daily.dev/posts/2YIPmtHiF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google Cloud launches N4A instances with Axion ARM CPUs</strong>: Google Cloud has released N4A compute instances powered by custom Axion ARM processors (Neoverse N3 cores), offering up to 64 vCPUs and 512GB RAM, targeting web servers, microservices, and containerized applications as a more cost-effective alternative to C4A instances. <a href="https://app.daily.dev/posts/JvBjVyZ5U" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Databricks SQL sees significant performance and cost improvements</strong>: Databricks SQL achieved up to 40% performance improvement in production workloads, 10x reduction in Unity Catalog latency, and 40% storage cost reduction via Zstandard compression in 2025, with all enhancements automatically applied in DBSQL Serverless. <a href="https://app.daily.dev/posts/rT9ZMYOFd" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Supabase launches AWS PrivateLink database connectivity in Beta</strong>: Supabase now offers PrivateLink in Beta for AWS users, enabling private network connections to Supabase databases via AWS VPC Lattice, enhancing security and compliance for regulated industries by eliminating public endpoints. <a href="https://app.daily.dev/posts/Bgvnt4bF8" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-driven polymorphic phishing uses LLMs for dynamic malware</strong>: A new phishing attack leverages AI models like DeepSeek and Gemini to generate malicious JavaScript client-side, splitting malware into fragments delivered from trusted LLM domains to evade traditional detection systems through ‘last mile reassembly.’ <a href="https://app.daily.dev/posts/jfmiKNwaJ" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>ChatGPT caught citing AI-generated Grokipedia: The echo chamber is real</strong>: The Guardian found ChatGPT’s GPT-5.2 model citing Elon Musk’s AI-generated encyclopedia, Grokipedia, raising concerns about recursive AI content amplifying bias and misinformation. OpenAI claims its models filter low-credibility sources, but apparently, an AI-generated encyclopedia passes the sniff test. So, we’re just building a bigger, faster echo chamber, where AI talks to AI, and the truth becomes whatever the algorithm decides. What could possibly go wrong? <a href="https://app.daily.dev/posts/fWRmbZwbT" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>D-Wave’s quantum computing claims: Growth, contracts, and a whole lot of self-promotion</strong>: D-Wave is touting 314% growth in Advantage2 usage and a $10 million QCaaS contract, while also claiming to be the only company offering all three core superconducting quantum technologies. All these impressive numbers and bold statements, however, come directly from D-Wave’s own press releases, with zero independent verification. It’s almost as if a company with a vested interest in looking good is, well, looking good. Color me shocked. <a href="https://app.daily.dev/posts/B9UXF3b8Z" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cursor’s AI browser: A ‘major breakthrough’ or just spaghetti code?</strong>: Cursor announced its AI agents built a web browser with millions of lines of code and a custom Rust rendering engine, hailing it as a ‘major AI software engineering breakthrough.’ However, actual developers who bothered to look at the code found it to be ‘spaghetti,’ unable to compile, and heavily reliant on existing projects, completely undermining the ‘from scratch’ claim. It seems the only thing ‘major’ here is the gap between marketing hype and reality. <a href="https://app.daily.dev/posts/gcgtwbk5H" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
