<h2>Must know</h2>
<ul>
<li><strong>Cursor 3 Glass demotes the IDE to a background process</strong>: Cursor has pivoted its entire architecture to treat the traditional code editor as a fallback for an agent management console. According to The New Stack, the new ‘Glass’ release features multi-repo workspaces and a Cloud Handoff system that syncs agent sessions between local and cloud environments. This shift signals a transition from writing code to orchestrating high-level agent workflows and reviewing diffs. <a href="https://app.daily.dev/posts/GsQ26ueQa" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Karpathy proposes the LLM wiki pattern for knowledge management</strong>: Andrej Karpathy has detailed a new architectural pattern for personal knowledge bases that moves beyond traditional RAG. Instead of querying raw docs, an LLM builds and maintains a structured, interlinked markdown wiki, handling the ingestion, contradiction flagging, and cross-referencing automatically. This approach, shared on Hacker News, aims to eliminate the manual maintenance burden that typically causes human-managed wikis to fail over time. <a href="https://app.daily.dev/posts/7o1BSKg72" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Alibaba releases Qwen 3.6 Plus for agentic coding</strong>: Alibaba’s Qwen team has launched Qwen 3.6 Plus, a closed-weights model specifically optimized for repo-level tasks and agentic reasoning. The model features a massive 1 million token context window and is being offered with a generous free tier of 1,000 requests per day via their open-source Qwen Code platform. While the weights are closed, its performance on practical coding benchmarks is reportedly rivaling top-tier flagship models. <a href="https://app.daily.dev/posts/UvbKi89vg" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Neovim 0.12.0 adds native autocomplete and a built-in plugin manager</strong>: The latest release reduces plugin bloat with native insert-mode completion and an experimental ‘vim.pack’ manager with lockfile support. <a href="https://app.daily.dev/feed-by-ids?id=3fbVLSQI2&amp;id=T9nUhWoBa" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Tinygrad eGPU drivers now run natively on macOS without SIP bypass</strong>: Apple has approved the tinygrad driver for AMD and NVIDIA eGPUs via Thunderbolt, simplifying ML acceleration on Mac hardware. <a href="https://app.daily.dev/posts/JhqNvWHDd" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>TurboQuant WASM brings 6x vector compression to the browser</strong>: A new experimental package uses relaxed SIMD to compress float vectors to 4.5 bits while preserving dot products for fast similarity search. <a href="https://app.daily.dev/posts/h7ZO3PZhH" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Storybook MCP integrates AI for component driven development</strong>: The new Model Context Protocol allows AI agents to generate UI features by directly consuming existing design system components. <a href="https://app.daily.dev/posts/WkXn6nL7A" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Roogle provides a type signature search engine for Rust</strong>: This open-source tool allows developers to query Rust APIs by function name or type signature via a local server or Docker. <a href="https://app.daily.dev/posts/L0uOWR5pB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Aegis Terra 1 is a fully open source FPGA from the silicon up</strong>: Unlike projects that wrap proprietary silicon, Aegis uses a Dart-based HDL to generate synthesizable SystemVerilog for the GF180MCU process. <a href="https://app.daily.dev/posts/9mxY3iV8A" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux 7.0 rc7 updates security reporting docs to handle AI spam</strong>: The kernel team is adding stricter documentation to filter out the recent surge of low-quality, AI-generated security bug reports. <a href="https://app.daily.dev/posts/SG3mIyq6B" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>The AI native professional identity crisis</strong>: C# Corner is pushing the idea of ‘AI-native’ as a new personality trait where you ‘operate through AI by default.’ It’s a classic rebranding of ‘using a tool’ into a fundamental shift in human evolution to justify corporate restructuring. Apparently, if you aren’t letting a LLM think for you 24/7, you’re just an ‘occasional user’ destined for the unemployment line. <a href="https://app.daily.dev/posts/6kbUCvBYV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NotebookLM’s cinematic whiteboard hallucinations</strong>: Google’s NotebookLM now generates ‘Cinematic Video Overviews’ that look like someone sketching on a whiteboard, which marketing claims ‘does what YouTube never could.’ In reality, it’s a $250/month feature that caps you at two videos a day because generating high-fidelity AI video is still prohibitively expensive and mostly just a visual gimmick for reading your own notes back to you. <a href="https://app.daily.dev/posts/QHfkwP567" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The case against Markdown’s existence</strong>: A recent critique argues Markdown is a security nightmare and a technical failure because it requires an HTML parser to function. While the ReDoS and XSS concerns are real, claiming the world’s most popular documentation format is ‘failing’ because it has dialects is the kind of pedantic gatekeeping that only makes sense if you’ve never had to write a README in LaTeX. It’s a document format, not a mission-critical kernel module. <a href="https://app.daily.dev/posts/DOoBXaOfV" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
