<h2>Must know</h2>
<ul>
<li><strong>Meta builds gigawatt-scale AI clusters with Prometheus</strong>: Meta is deploying gigawatt-scale AI clusters, codenamed Prometheus, using a centralized Ethernet-based super spine network called Backend Aggregation (BAG). This network interconnects multiple data center fabrics across regions, supporting petabit-range bandwidth between tens of thousands of GPUs. The architecture leverages modular Jericho3 ASIC chassis and eBGP with UCMP for load balancing, enabling massive AI compute capacity across multiple data center buildings, as reported by Facebook Engineering. <a href="https://app.daily.dev/posts/L6vqc64qy" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Heroku is officially dead, time to migrate</strong>: Salesforce has moved Heroku into a ‘sustaining engineering’ phase, halting new enterprise contract sales and effectively signaling the platform’s end. This means no new features and a clear path towards obsolescence. Existing customers are advised to treat Heroku as legacy infrastructure and plan migrations to alternatives like Render, <a href="http://Fly.io" target="_blank" rel="noopener nofollow">Fly.io</a>, or Railway, 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=pW1DiQ44D&amp;id=smfFrov84" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux kernel permanently adopts Rust</strong>: Linux 7.0 officially concludes the ‘Rust experiment’ by making Rust a permanent part of the kernel project. This decision signals a long-term commitment to the language, encouraging investment in Rust training and development within the kernel community. The move is a culmination of years of support and ongoing improvements, solidifying Rust’s role in future kernel development. <a href="https://app.daily.dev/posts/k0qXAqGCH" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>NVIDIA automates LLM inference optimization with TensorRT LLM AutoDeploy</strong>: NVIDIA’s new TensorRT LLM AutoDeploy automates the compilation of PyTorch models into optimized inference engines for LLMs, supporting over 100 text-to-text models and achieving production-ready performance in days. <a href="https://app.daily.dev/posts/QkLXaUNML" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>North Korean threat actors use AI deepfakes in crypto attacks</strong>: North Korean threat actor UNC1069 is using AI-generated deepfake video and compromised Telegram accounts in sophisticated attacks targeting the cryptocurrency sector, deploying new malware to harvest credentials and wallet information. <a href="https://app.daily.dev/posts/wgv654vDT" 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 undergoing major performance improvements in versions 8 and 9, with redesigned hash table architecture, reduced memory overhead, and up to 40% memory usage reduction for small key-value workloads while maintaining throughput. <a href="https://app.daily.dev/posts/BJPWCn9Zl" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google Cloud Monitoring enhances OTLP metrics support</strong>: Google Cloud Monitoring now supports OpenTelemetry Protocol (OTLP) for metrics, enabling provider-agnostic telemetry pipelines, direct SDK-to-Cloud Monitoring ingestion, and zero-code auto-instrumentation. <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, allowing modifications to existing machines without recreation and enabling skipping multiple Kubernetes minor versions in a single operation. <a href="https://app.daily.dev/posts/SwlgWxJBR" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Black Basta ransomware integrates BYOVD for EDR evasion</strong>: The Black Basta ransomware gang is now bundling a vulnerable driver (NsecSoft NSecKrnl) directly into its payload to terminate EDR solutions and gain kernel-level access, speeding up attacks and reducing detection chances. <a href="https://app.daily.dev/posts/xkkTWSdYu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New DNS-PERSIST-01 validation method simplifies certificate automation</strong>: A new DNS-PERSIST-01 domain control validation method is coming in 2026, allowing CAs to issue certificates based on a single, static DNS record, significantly reducing complexity for certificate automation. <a href="https://app.daily.dev/posts/7hFHkMPwg" 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, but it’s terrible</strong>: Anthropic researchers used 16 Claude Opus 4.6 AI agents to build a Rust-based C compiler that successfully compiles the Linux kernel. The catch? It cost $20,000 in API calls, generated 100,000 lines of subpar Rust, 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 useful as a screen door on a submarine for production. <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’s Claude Co-work plugin ‘wiped out $400 billion in tech stocks’</strong>: Anthropic released Claude Co-work with industry-specific plugins, and some sources are claiming this ‘wiped out $400 billion in tech stocks.’ Apparently, a new AI tool is enough to send the entire SaaS market into a tailspin. Or maybe, just maybe, market concerns and questions about future disruption aren’t quite the same as a definitive, massive, direct loss. But hey, why let nuance get in the way of a good headline? <a href="https://app.daily.dev/posts/QnkkrfiEd" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI therapy chatbots are here to solve all your mental health problems</strong>: MIT Technology Review reports on the rise of AI therapy chatbots like Wysa and Woebot, suggesting they’re addressing mental health needs. Meanwhile, an Oxford study found that AI chatbots don’t improve medical advice compared to search engines and give inconsistent, even contradictory, recommendations. So, if you’re looking for a therapist who might tell you to do the exact opposite of what they just said, AI’s got you covered! <a href="https://app.daily.dev/posts/jmvy6d70v" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
