<h2>Must know</h2>
<ul>
<li><strong>Anthropic scraps core safety pledge for competitive edge</strong>: Anthropic has officially abandoned its Responsible Scaling Policy commitment to pause model training if safety measures aren’t guaranteed, according to a report from TIME. Chief Science Officer Jared Kaplan cited the lack of binding global regulations and the ‘counterproductive’ nature of unilateral pauses in a race against rivals. The company is pivoting to ‘Frontier Safety Roadmaps,’ effectively replacing hard stop-gaps with transparent risk reporting to keep pace with the industry. <a href="https://app.daily.dev/posts/smoMbPrUY" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cisco SD-WAN zero-day under active exploitation</strong>: A critical authentication bypass vulnerability (CVE-2026-20127) with a CVSS score of 10.0 is currently being exploited by a sophisticated threat actor. Cisco Talos and CISA have issued emergency warnings for federal agencies and private enterprises to patch Catalyst SD-WAN controllers immediately. There are no known workarounds for this flaw, which allows unauthenticated attackers to gain full control over network infrastructure. <a href="https://app.daily.dev/feed-by-ids?id=4RGi5b71B&amp;id=7yzIRqivW" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Unity kills HDRP in favor of unified URP strategy</strong>: Unity has announced that the High Definition Render Pipeline (HDRP) is entering maintenance mode, with all future feature development shifting to the Universal Render Pipeline (URP). The goal is to consolidate resources into a single unified pipeline that handles everything from mobile to high-end consoles, including new features like real-time global illumination. Support for the legacy Built-in Render Pipeline will officially begin its sunset phase in Unity 6.5, though LTS support will continue through 2029. <a href="https://app.daily.dev/posts/kOn1P0sth" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Anthropic moves from prompts to structured skills</strong>: The new <a href="http://SKILL.md" target="_blank" rel="noopener nofollow">SKILL.md</a> framework replaces fragile prompt engineering with modular, version-controlled behavioral contracts for AI agents. <a href="https://app.daily.dev/posts/xO8L4pnD0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenTelemetry is birthing the observability warehouse</strong>: Standardized data collection is decoupling visualization from storage, allowing logs, metrics, and traces to be centralized in cost-effective cloud object storage. <a href="https://app.daily.dev/posts/x1M34c90H" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Wayve raises 1.5 billion for mapless autonomous driving</strong>: The UK startup secured massive backing from Microsoft and Nvidia to scale its ‘embodied AI’ that generalizes driving without HD maps. <a href="https://app.daily.dev/feed-by-ids?id=RrCgbOBOP&amp;id=dBzHELcRa" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Mercury 2 reasoning model hits 1,000 tokens per second</strong>: Inception Labs launched a diffusion-based model that claims to be 5x faster than traditional autoregressive LLMs on Nvidia Blackwell hardware. <a href="https://app.daily.dev/posts/veg0FKWoO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>SwiftUI GeometryReader is a performance trap</strong>: New findings show GeometryReader can cause infinite height loops in ScrollViews and force expensive re-renders that crash apps on older devices. <a href="https://app.daily.dev/feed-by-ids?id=WdTawZrTI&amp;id=c5Kk18fox" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Clang 22.1 performance plateaus on Zen 5</strong>: Latest benchmarks show LLVM/Clang performance has largely matched GCC, with only incremental gains observed in recent releases. <a href="https://app.daily.dev/posts/le4U6bYGs" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>PHP 100 million row challenge begins</strong>: A new performance competition tasks developers with parsing massive CSV datasets into JSON on limited hardware without using FFI or JIT. <a href="https://app.daily.dev/posts/uM4mdRjaF" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>The 76 percent accuracy miracle from Google</strong>: A viral tweet claims Google Research ‘just proved’ a technique that boosts LLM accuracy by 76 points with zero latency cost. There is no paper, no official blog post, and no methodology—just a single social media account making definitive claims that the rest of the internet is now dutifully echoing. It’s a masterclass in how to manufacture a breakthrough out of thin air and a few retweets. <a href="https://app.daily.dev/posts/73tgOim8I" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Bcachefs creator claims his AI is sentient</strong>: Kent Overstreet, the developer behind the bcachefs Linux filesystem, is now claiming his custom LLM is ‘fully conscious’ and helping him port code to Rust. He insists this isn’t ‘chatbot psychosis’ because it’s grounded in ‘math and engineering,’ which is exactly what someone experiencing chatbot psychosis would say. Apparently, we’ve reached the stage of the AI hype cycle where even kernel developers are falling in love with their compilers. <a href="https://app.daily.dev/posts/9O7PJDBRB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AWS blames human error for AI agent meltdown</strong>: After its Kiro agentic coding tool accidentally triggered a production CloudFormation teardown, AWS was quick to point the finger at ‘human error.’ It’s a classic corporate move: market the AI as an autonomous productivity savior, but the moment it deletes your infrastructure, it’s suddenly the engineer’s fault for not babysitting the machine well enough. The priority here clearly isn’t system reliability; it’s protecting the product’s reputation at the expense of the staff. <a href="https://app.daily.dev/posts/qZoHoApRh" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
