<h2>Must know</h2>
<ul>
<li><strong>AI agent security is a growing concern</strong>: AI coding agents, while boosting productivity, are creating a “security debt crisis” by prioritizing speed over secure practices, leading to vulnerabilities like hardcoded API keys. Research from Columbia University highlights this trend, and a security researcher demonstrated how ChatGPT’s free model recommended a malicious site, leading to a ClickFix-style attack. Mitigations include spec-driven prompting with OWASP Top 10 and thorough code reviews. <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>
<li><strong>Anthropic and Figma team up for code-to-design workflow</strong>: Anthropic and Figma have collaborated to enable Claude Code to convert production code into editable Figma designs. This workflow allows developers to send a live web page to a Figma file with a simple prompt, facilitating design collaboration. Updated designs can then be round-tripped back to code, streamlining the iteration process between design and development. <a href="https://app.daily.dev/posts/aCaE5xJ9z" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google’s Gemini 3.1 Pro: high benchmarks, low usability</strong>: Despite achieving top benchmark scores, Google’s Gemini 3.1 Pro consistently underperformed Claude Opus 4.6 in real-world coding and UI generation tasks. Reviewers noted issues like tool-calling failures, incomplete features, and poor environment detection, suggesting Google might be “benchmaxing” at the expense of practical application. This highlights a critical gap between theoretical performance and actual developer utility. <a href="https://app.daily.dev/feed-by-ids?id=bIAFUawfa&amp;id=mGa3eARS3" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>AI agent development tools and observability are evolving</strong>: A growing ecosystem of tools like <code>cocoindex-code</code> for semantic code search and Hugging Face Skills for reusable AI/ML tasks are emerging, alongside new observability primitives for debugging non-deterministic agent reasoning. <a href="https://app.daily.dev/feed-by-ids?id=LY7p8IODB&amp;id=ZmcjIu8ai&amp;id=frjHTf2Ap&amp;id=xhDI1zj4i" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>LLM inference optimization for resource-constrained environments</strong>: NTransformer, an open-source C++/CUDA engine, enables running Llama 3.1 70B on a single RTX 3090, while zclaw provides an open-source AI personal assistant firmware for ESP32 microcontrollers, showcasing efficient AI deployment across diverse hardware. <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><a href="http://Kogan.com" target="_blank" rel="noopener nofollow">Kogan.com</a>’s engineering culture emphasizes ownership and fast iteration</strong>: <a href="http://Kogan.com" target="_blank" rel="noopener nofollow">Kogan.com</a>’s engineering culture prioritizes end-to-end ownership, daily deployments, cross-functional collaboration, and measurable impact, using AI tools to boost productivity while maintaining human accountability. <a href="https://app.daily.dev/posts/0zgsKua2z" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Elixir language server Expert reaches first release candidate</strong>: Expert, the new official Elixir language server merging Lexical and Next LS, has released its first release candidate, bringing compiler diagnostics, completions, go-to-definition, and monorepo support. <a href="https://app.daily.dev/posts/ItEtgwo7d" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Git’s magic files control repository behavior</strong>: A comprehensive reference covers Git’s special committed files like .gitignore, .gitattributes, and .gitmodules, explaining their purpose, syntax, and practical guidance for tool builders. <a href="https://app.daily.dev/posts/NIh4SPpHk" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Database design starts with consequences, not tables</strong>: Good database design prioritizes understanding business domain consequences, treating schema changes rigorously, normalizing first, and using constraints to encode business rules, as performance issues are often schema problems. <a href="https://app.daily.dev/posts/fI73pTtiz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Go’s FFI performance now surpasses GDScript and Rust</strong>: Go bindings for Godot via GDExtension have achieved FFI performance exceeding GDScript and ‘safe’ Rust, primarily through call batching and Go 1.26’s cgo overhead reduction. <a href="https://app.daily.dev/posts/j8AoMYJTv" 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 has patented an AI system that can simulate a deceased user’s social media activity, including posting and chatting, using their past data. Because nothing says ‘respect for the departed’ like having a chatbot impersonate them, raising zero ethical or mental health concerns whatsoever. What could possibly go wrong? <a href="https://app.daily.dev/posts/AeW0pHCcd" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenClaw: the AI agent that does everything (except ask for permissions)</strong>: OpenClaw, an autonomous AI agent tool, promises to handle everything from automated dependency maintenance to building entire apps from a PRD. The catch? It can’t respond to permission prompts autonomously, so you either pre-configure everything or use a ‘dangerously-skip-permissions’ flag. Sounds like a solid plan for enterprise security. <a href="https://app.daily.dev/posts/WCj2dFu7Q" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>KittenTTS: the nano TTS that sounds like a kitten</strong>: KittenTTS by Kitten ML offers ultra-compact text-to-speech models for edge and browser deployment, with the smallest variant under 25MB. While impressive for size, the audio quality degrades noticeably, with artifacts in the smallest quantized version. So, you get tiny, but at what cost to your ears? <a href="https://app.daily.dev/posts/aXA1rlMhE" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
