<h2>Must know</h2>
<ul>
<li><strong>Google launches Gemma 4 with native multimodality</strong>: Google released the Gemma 4 open-weight model family under Apache 2.0, featuring native support for vision, audio, and 140+ languages. The lineup includes a 26B MoE model for low-latency workstation use and smaller edge variants optimized for Android’s AICore and Raspberry Pi 5. According to Google DeepMind and NVIDIA, these models support a 256K context window and are immediately available via Hugging Face, Ollama, and vLLM. <a href="https://app.daily.dev/feed-by-ids?id=DxUdrK5VU&amp;id=hi4RmWdI4&amp;id=hUcJhitIB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic source code leak exposes internal agent logic</strong>: A misconfigured npm package accidentally leaked 512,000 lines of Claude Code’s TypeScript source code via a public source map file. The leak, confirmed by reports from Axios and Dev Weekly, revealed unreleased features like ‘KAIROS’ for persistent autonomous agents and an ‘Undercover Mode’ that strips AI attribution from git commits. While Anthropic has issued thousands of DMCA takedowns, the architectural blueprints for their agentic orchestration are already circulating in the community. <a href="https://app.daily.dev/feed-by-ids?id=9nHqJG8ou&amp;id=coOVHaDO4&amp;id=CGbQv469o&amp;id=lSeOtUf5e" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Meta introduces semi-formal reasoning for code verification</strong>: Meta researchers have developed a ‘semi-formal reasoning’ technique that allows AI agents to verify code correctness without execution by constructing explicit premises and tracing paths. This method reportedly improved patch equivalence verification accuracy from 78% to 93% and boosted fault localization scores on Defects4J. Meta suggests this could significantly reduce the need for expensive sandbox-based reinforcement learning pipelines in DevOps workflows. <a href="https://app.daily.dev/posts/gfdiiQquJ" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Pretext library enables DOM-less text height calculation</strong>: React core dev Cheng Lou released a minimal JS library that uses off-screen canvas and word-wrap emulation for high-performance dynamic layouts. <a href="https://app.daily.dev/posts/rS0vBup8W" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Intl API remains the best npm-free formatting solution</strong>: The native browser Intl API supports 7,000+ locales for dates, currencies, and list formatting, replacing heavy dependencies like date-fns or Moment.js. <a href="https://app.daily.dev/posts/ZFYf8vz1W" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Vercel Sandbox restores now hit sub-second speeds</strong>: Engineering optimizations involving parallelized S3 range requests and NVMe-backed LRU caching reduced p75 filesystem restore times from 40s to under 1s. <a href="https://app.daily.dev/posts/1Xib7hCkG" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>PHP 8.5 features and ecosystem updates</strong>: The ‘Can I PHP?’ reference now tracks PHP 8.5 features like the pipe operator, while Laravel 13.3 adds a #[UnitTest] attribute for faster, boot-less testing. <a href="https://app.daily.dev/feed-by-ids?id=Aa7VPPhsH&amp;id=T6Ccxev8J" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Model Context Protocol gains Postgres and CMS support</strong>: New MCP servers from pgEdge and Waldium allow AI agents to query databases and CMS content directly with 30-50% lower token usage via TSV output. <a href="https://app.daily.dev/feed-by-ids?id=09ZKKSp9j&amp;id=sCEi0lM0l" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Jujutsu VCS v0.40 adds diff-based revset functions</strong>: The Git-compatible tool now supports matching commits by lines added or removed and features a smarter TUI for arranging commit hierarchies. <a href="https://app.daily.dev/posts/XNj5BzBCW" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>View Transitions Toolkit simplifies animation orchestration</strong>: A new utility package provides helpers for feature detection, playback control, and automatic transition type injection for the View Transition API. <a href="https://app.daily.dev/posts/UTG40VF0N" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>The junior death spiral narrative</strong>: Microsoft execs and various ‘thought leaders’ are sounding the alarm on a 13% drop in junior hiring, blaming an ‘AI drag’ that makes entry-level roles obsolete. While they push ‘preceptor programs’ as a cure, the reality is that companies are simply using AI as a convenient excuse for financially-motivated headcount reductions. It’s less about ‘narrowing the pyramid’ and more about executives ‘AI washing’ their quarterly budget cuts. <a href="https://app.daily.dev/feed-by-ids?id=Qfe9o2iyU&amp;id=KVQvWvGP4&amp;id=uIab7G9MK" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Upscaling as a substitute for optimization</strong>: The PC gaming industry has officially given up on performance, with developers now treating DLSS and FSR as mandatory crutches rather than optional enhancements. Reports indicate that even high-end hardware struggles without software-generated frames, while GPU makers pivot their best silicon to AI data centers. We’re being sold ‘proprietary solutions’ like DLSS 5 as innovation when they’re actually just masking lazy, unoptimized engine code. <a href="https://app.daily.dev/feed-by-ids?id=M5vdbYsqM&amp;id=Vs0fd08bp" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The exact blueprint for senior developer wealth</strong>: A series of identical articles are peddling an ‘exact blueprint’ to become a highly-paid senior dev by using the ‘RAG Matrix’ for personal KPIs. This is classic career-hustle spam that suggests shipping code is ‘insufficient’ and that you should spend your time generating ‘evidence of impact’ instead. If you’re spending more time on your personal promotion dashboard than on your actual architecture, you’re not a senior engineer; you’re a middle manager in training. <a href="https://app.daily.dev/feed-by-ids?id=eWWWY2wcm&amp;id=uAKINimfh&amp;id=l3RTOH4WS" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
