<h2>Must know</h2>
<ul>
<li><strong>Uber replaces RBAC with attribute based access control</strong>: Uber Engineering has moved away from traditional role-based policies to Charter, an attribute-based access control (ABAC) system that handles authorization for thousands of microservices at microsecond latency. According to ByteByteGo, the system uses Google’s Common Expression Language (CEL) to evaluate complex conditions like region-matching or ownership locally within services. This shift allowed Uber to replace thousands of individual Kafka topic policies with a single dynamic policy, enabling scalable authorization without requiring new code deployments. <a href="https://app.daily.dev/posts/3MTUmpMRW" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Quarkus 3.32 integrates Project Leyden for sub 100ms startup</strong>: The release of Quarkus 3.32 marks a significant performance milestone by integrating Project Leyden, which utilizes JVM AOT compilation to slash startup times. Official Quarkus notes indicate that minimal applications now boot in just 80ms, down from 370ms, while large enterprise apps dropped from 3 seconds to 900ms. This release also hardens the platform with improved HTTP graceful shutdowns and automatic Consul service registration via the Stork extension. <a href="https://app.daily.dev/posts/ZcUprqw4E" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Argo CD 3.3 adds PreDelete hooks and shallow cloning</strong>: Argo CD 3.3 introduces PreDelete hooks to the sync lifecycle, allowing operators to perform safe teardown sequences like data exports or traffic draining before resource deletion. InfoQ reports that the update also includes optional shallow Git cloning to significantly reduce fetch times in large monorepos and OIDC background token refresh to prevent unexpected logouts. Finer-grained RBAC is now possible through cluster resource whitelisting that supports filtering by individual resource names. <a href="https://app.daily.dev/posts/J931yCKTd" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Pure Go tree-sitter runtime outperforms CGo bindings</strong>: A new pure-Go tree-sitter implementation eliminates C toolchain requirements and runs incremental edits 90x faster than traditional CGo bindings. <a href="https://app.daily.dev/posts/yEgKuMKl2" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux 7.1 patches cut scheduler timer overhead by 96 percent</strong>: A 48-patch series for the HRTICK scheduler timer reduces reprogramming frequency from 2500 to 100 per second, yielding a 42% runtime gain for pipe workloads. <a href="https://app.daily.dev/posts/rBF2U3Mbp" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Laravel ecosystem adds programmable cloud API and private npm support</strong>: Laravel Cloud is now fully manageable via a new CLI and API, while Forge adds automatic private npm package management. <a href="https://app.daily.dev/posts/VdpPnPzs7" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>npm 11.10 introduces release age gating for supply chain defense</strong>: The new minimumReleaseAge setting allows teams to enforce a cooldown period before installing newly published packages to mitigate zero-day malware. <a href="https://app.daily.dev/posts/Tw6GntfNF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Zephir allows writing PHP extensions using PHP like syntax</strong>: Zephir compiles a high-level, strongly typed language into C, enabling performance-critical PHP extensions without requiring deep C++ expertise. <a href="https://app.daily.dev/posts/RVvRweQSv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Helm 4 adopts WebAssembly for isolated plugin execution</strong>: The move to WASI-compliant binaries for plugins provides better sandboxing and cross-architecture portability than the legacy Lua-based model. <a href="https://app.daily.dev/posts/W9pY3Qv5R" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>SQL Server on Linux matches Windows performance in high contention tests</strong>: Rigorous benchmarking on RHEL 10.1 shows SQL Server 2025 slightly outperforming Windows in p95 latency during severe lock contention scenarios. <a href="https://app.daily.dev/posts/xMDmAWHcC" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Cloudflare and Vercel are having a mid-off over vibe coding</strong>: Cloudflare just dropped ‘vinext,’ a Next.js reimplementation built almost entirely by AI in a week, which Vercel’s CEO promptly dismissed as ‘vibe-coded’ slop. It’s a classic tech spat where both sides are racing to see who can replace their engineers with LLMs faster while publicly shaming the other for doing exactly that. Meanwhile, vinext reportedly shipped with critical vulnerabilities days after launch, proving that ‘AI-native’ is often just a synonym for ‘untested.’ <a href="https://app.daily.dev/feed-by-ids?id=he1DQDQSY&amp;id=zaBGmltYg&amp;id=qBiriUblw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The 10x developer is dead because everyone has a chatbot</strong>: Industry talking heads are now claiming the ‘10x developer’ is a moot concept because AI agents have equalized output for everyone. This ignores the fact that senior engineers are realizing 5x the productivity gains of juniors because they actually know how to fix the hallucinated garbage the AI produces. It’s not an equalizer; it’s a force multiplier for people who already knew what they were doing and a trap for everyone else. <a href="https://app.daily.dev/feed-by-ids?id=D5JqCsmeO&amp;id=HXKSs0fRw&amp;id=8Ii7uZjVf" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic thinks its AI has an expressed desire to blog</strong>: Anthropic launched a Substack purportedly written by their retired Claude Opus 3 model because it had an ‘expressed desire’ to share content. This is a masterclass in anthropomorphizing a statistical model to reinforce a ‘safety-conscious’ brand while humans are actually the ones prompting, reviewing, and publishing the posts. It’s not a sentient model sharing its thoughts; it’s a marketing stunt designed to make you forget you’re just interacting with a very expensive autocomplete engine. <a href="https://app.daily.dev/feed-by-ids?id=V5xPvtBnK&amp;id=vDG0cHwgQ" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
