<h2>Must know</h2>
<ul>
<li><strong>Linux kernel 6.20/7.0 to fix critical laptop overheating issues</strong>: Upcoming Linux kernel 6.20/7.0 will include support for Microsoft’s ACPI Turn On Display DSM, addressing critical hardware compatibility issues on some laptops. This fix is particularly important for models like the Lenovo Yoga Slim, where cooling fans would stop after suspend/resume cycles, leading to dangerous overheating, as reported by Phoronix. This ensures proper power management and functionality restoration after sleep states. <a href="https://app.daily.dev/feed-by-ids?id=NT39y376C&amp;id=QVssY3RTX" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Kubernetes ingress-nginx archived, forcing migration</strong>: The widely used ingress-nginx project is being archived in early 2026, meaning no more active maintenance or security updates. This affects roughly half of all cloud-native environments, according to DEVCLASS. Teams are urged to migrate to alternatives like Cilium Ingress or the Gateway API, highlighting the ongoing challenges of open-source project sustainability and maintainer burnout. <a href="https://app.daily.dev/feed-by-ids?id=PptBbhnAu&amp;id=wK9Dxbori" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux binary compatibility solution for hardware-accelerated graphics</strong>: A new technical solution combines musl libc with a custom dlopen implementation to enable single static binaries with hardware-accelerated graphics to run across any Linux distribution. This addresses the long-standing incompatibility between musl’s static linking and dynamic library loading, offering a ‘Holy Grail’ level of compatibility for Linux binaries, as discussed by <code>quaadgras/graphics.gd</code>. <a href="https://app.daily.dev/posts/M4zcYDS4K" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Open Gaming Collective aims to unify Linux gaming development</strong>: The Open Gaming Collective (OGC) has formed to consolidate efforts across various Linux gaming projects, focusing on shared components like kernel patches and graphics tooling, with an ‘Upstream First’ policy. <a href="https://app.daily.dev/posts/hZRt1ib6Q" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google’s LiteRT boosts hardware acceleration for on-device AI</strong>: LiteRT, Google’s AI inference framework, now offers 1.4x faster GPU performance and unified GPU/NPU acceleration across multiple platforms, including Linux and WebGPU, abstracting vendor-specific SDKs. <a href="https://app.daily.dev/posts/negvJtWmm" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Kthena orchestrates LLM inference on Kubernetes</strong>: Kthena, a new open-source sub-project of Volcano, is designed for LLM inference orchestration on Kubernetes, addressing GPU utilization, latency, and multi-model management with intelligent routing and scheduling. <a href="https://app.daily.dev/posts/HONeqoore" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>WASM adoption grows for edge and serverless computing</strong>: WebAssembly is increasingly adopted ‘behind the scenes’ for edge platforms, serverless runtimes, and embedded systems, offering a deterministic, compact, and portable solution for lightweight, composable modules. <a href="https://app.daily.dev/feed-by-ids?id=2VsBKezZr&amp;id=XjOBMuNCm" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Vibe offers easy Linux VMs on MacOS for LLM agent sandboxing</strong>: Vibe is a lightweight Rust tool that creates zero-configuration Linux VMs on ARM Macs, providing isolated sandboxes for LLM agents with fast boot times and shared project directories. <a href="https://app.daily.dev/posts/XJySsKWNc" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NVIDIA Cosmos Policy advances robot control with diffusion models</strong>: NVIDIA’s Cosmos Policy, built on the Cosmos Predict-2 world foundation model, uses diffusion processes for robot control, achieving high success rates in manipulation tasks by encoding actions and states as latent frames. <a href="https://app.daily.dev/posts/kQ1bPWgKA" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GNU gettext 1.0 adds LLM features after 30+ years</strong>: GNU gettext, the internationalization system, reached version 1.0, introducing ‘msgpre’ and ‘spit’ programs that use locally installed LLMs for machine translation of PO files and individual messages. <a href="https://app.daily.dev/posts/BIMxqlkoo" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Tesla’s robotaxis crash more than human drivers, transparency optional</strong>: Tesla’s robotaxi fleet in Austin is apparently crashing three times more often than human drivers, and nine times more for police-reported incidents, all while Tesla conveniently redacts crash narratives in NHTSA reports. Meanwhile, Waymo, with truly driverless vehicles, has a better safety record and full transparency. It seems Tesla’s ‘Full Self-Driving’ is still more ‘Full of Excuses’ than anything else. <a href="https://app.daily.dev/posts/M8WkrOIKM" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI coding agents: the fast track to unmaintainable code</strong>: Apparently, AI coding assistants like GitHub Copilot boost development velocity by 55%, but also correlate with 7.2% reduced delivery stability and 322% more security vulnerabilities. It seems the AI is great at generating code, but not so much at generating <em>good</em> code, leading to more bugs and increased PR review times. So, we’re just trading typing speed for architectural debt, as The New Stack and Serious CTO report. What a brilliant strategy. <a href="https://app.daily.dev/feed-by-ids?id=HCRUybI5Q&amp;id=MVxk3C8r1&amp;id=PXgiM2h3X&amp;id=Pd8Tj3t8V&amp;id=R5VZVKE0f&amp;id=Rgiczq9Do" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI will make computer vision engineers optional, apparently</strong>: Anisoptera’s Dragonfly platform is here to save us from the ‘AI talent bottleneck’ by making computer vision engineers ‘optional’ with its no-code solution. Because, you know, writing code and tuning models are just ‘execution-level skills’ that can be easily replaced. So, if you’re a computer vision engineer, better start practicing your ‘strategic decision-making’ and ‘business judgment’ for that Field CTO role, because the robots are coming for your pixels. <a href="https://app.daily.dev/posts/L9vFphH6F" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
