<h2>Must know</h2>
<ul>
<li><strong>Anthropic and Figma bridge code to design</strong>: Anthropic and Figma teamed up to create a workflow that converts production code into editable Figma designs using Claude Code. This allows developers to send a live web page to Figma with a simple prompt, facilitating design collaboration and enabling round-trips back to code. It’s a pretty neat integration for bridging the design-to-code gap, as announced on February 21, 2026. <a href="https://app.daily.dev/posts/aCaE5xJ9z" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Gemini 3.1 Pro: high benchmarks, low usability</strong>: Google’s Gemini 3.1 Pro, despite boasting impressive benchmark scores like 78% on ARC AGI 2, is reportedly a letdown in real-world coding and UI generation tasks. Multiple sources, including articles from February 21-22, 2026, highlight issues like tool-calling failures, incomplete features, and poor environment detection, making it consistently underperform against competitors like Claude Opus 4.6 in practical applications. It seems Google might be ‘benchmaxing’ at the expense of actual usability. <a href="https://app.daily.dev/feed-by-ids?id=bIAFUawfa&amp;id=mGa3eARS3" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI coding agents are creating security debt</strong>: AI coding agents, while boosting productivity, are creating a ‘security debt crisis’ by prioritizing speed over secure practices, according to Columbia University research. Issues like hardcoded API keys and public access policies are becoming common. Mitigations include spec-driven prompting with OWASP Top 10 and thorough code reviews, but the trend suggests a need for more robust security measures in AI-driven development workflows. <a href="https://app.daily.dev/feed-by-ids?id=DD6LZnw0Q&amp;id=Et2d8uHu3&amp;id=LUO8dPKde&amp;id=RH1TYGIET&amp;id=UGo4Ea1t8&amp;id=bDJrJKMJX&amp;id=em7afQ5sm&amp;id=l0KLAjYGc&amp;id=ohhVPexbM&amp;id=vXBWwcwBk" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Google’s WebMCP enables browser-native AI agents</strong>: WebMCP is a new browser-native API from Google that allows AI agents to invoke structured JavaScript tools directly within the browser, improving determinism and reducing token overhead. <a href="https://app.daily.dev/posts/61GmU0Tcy" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NTransformer optimizes LLM inference on consumer GPUs</strong>: NTransformer, an open-source C++/CUDA engine, claims to run Llama 3.1 70B on a single RTX 3090 by streaming layers and using speculative decoding, achieving significant speedups. <a href="https://app.daily.dev/feed-by-ids?id=Ahx4zgQks&amp;id=WoJMuyqAQ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>LangSmith Agent Builder uses filesystem for agent memory</strong>: LangSmith Agent Builder uses a filesystem-based memory system, storing agent knowledge in Postgres and exposing it as a virtual filesystem, allowing agents to update their own memory. <a href="https://app.daily.dev/posts/9xLrRFK2b" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Database design starts with consequences, not tables</strong>: Good database design starts with understanding business consequences, emphasizing schema changes, deliberate denormalization, proper indexing, and using constraints to encode business rules. <a href="https://app.daily.dev/posts/fI73pTtiz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong><a href="http://Kogan.com" target="_blank" rel="noopener nofollow">Kogan.com</a>’s engineering culture focuses on ownership and speed</strong>: <a href="http://Kogan.com" target="_blank" rel="noopener nofollow">Kogan.com</a>’s engineering culture emphasizes strong ownership, fast iteration with daily deployments, cross-functional collaboration, and using AI tools while maintaining engineer accountability. <a href="https://app.daily.dev/posts/0zgsKua2z" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Adobe and NVIDIA achieve real-time sparkle rendering</strong>: A new research technique from Adobe and NVIDIA enables real-time rendering of millions of microscopic glint particles at over 280 FPS on consumer GPUs by generating particle positions on-the-fly. <a href="https://app.daily.dev/posts/rT496vKCA" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong><a href="http://Lobste.rs" target="_blank" rel="noopener nofollow">Lobste.rs</a> migration to SQLite hits performance snag</strong>: <a href="http://Lobste.rs" target="_blank" rel="noopener nofollow">Lobste.rs</a> attempted to migrate from MariaDB to SQLite but quickly reverted due to severe performance degradation, with plans to retry after addressing the issues. <a href="https://app.daily.dev/posts/RLQeRSeS0" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Meta patents AI to post after you’re dead</strong>: Meta apparently got a patent for an AI system that can simulate a deceased user’s social media activity, including posting and chatting, using their past data. Because what we all really need is our dead relatives’ AI ghosts haunting our feeds, courtesy of Meta. Nothing creepy about that at all. <a href="https://app.daily.dev/posts/AeW0pHCcd" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Orbital compute is the new AI battleground</strong>: Apparently, the global AI race is now all about ‘orbital compute’ and ‘infrastructure control,’ with SpaceX-xAI leading the charge with rockets and satellite constellations. Because terrestrial AI is just too slow and grid-constrained, we need AI in space. Sure, let’s just ignore the actual algorithmic challenges and focus on who has the biggest space toys. <a href="https://app.daily.dev/posts/QqzFDNqqS" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Breaking into tech via tech support</strong>: Some folks are still pushing the idea that you can break into tech without a CS degree by starting as a Tech Support Rep. While it’s true that internal mobility exists, let’s not pretend it’s a guaranteed fast track to CTO. It’s a grind, and ‘basic digital tool fluency’ isn’t quite the same as system design expertise. <a href="https://app.daily.dev/posts/IPyLPRMat" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
