<h2>Must know</h2>
<ul>
<li><strong>AI agent security is a mess</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 and public access policies. Research from Columbia University highlights this, and a security researcher demonstrated how ChatGPT recommended a malicious site for Tesseract OCR, leading to a reverse C2 shell. 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 and enabling updated designs to be round-tripped back to code. This could significantly streamline the design-to-development process, 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>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 utility for developers. <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 orchestration with Mission Control</strong>: Mission Control is an open-source Next.js 14 and TypeScript dashboard for orchestrating AI agents, managing tasks, and tracking multi-agent collaboration. <a href="https://app.daily.dev/posts/vyr9dJRQq" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NTransformer runs Llama 3.1 70B on a single RTX 3090</strong>: NTransformer, an open-source C++/CUDA LLM inference engine, can run Llama 3.1 70B on a single 24GB RTX 3090 by streaming layers via PCIe and using a 3-tier adaptive caching system. <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>WebMCP: browser-native AI agent execution</strong>: Google’s WebMCP is a browser-native API allowing websites to expose structured JavaScript tools for AI agents to invoke 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>KittenTTS: ultra-compact text-to-speech models</strong>: KittenTTS by Kitten ML offers a family of ultra-compact text-to-speech models (down to 15M parameters) optimized for edge and browser deployment, released under Apache 2.0. <a href="https://app.daily.dev/posts/aXA1rlMhE" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agent memory system in LangSmith Agent Builder</strong>: LangSmith Agent Builder uses a filesystem-based memory system, storing agent knowledge as files in Postgres and exposing it as a virtual filesystem, allowing agents to evolve their understanding across sessions. <a href="https://app.daily.dev/posts/9xLrRFK2b" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Kotlin Weekly #499 highlights new libraries</strong>: Kotlin Weekly #499 features new libraries like ProjectGuard for architectural boundary enforcement, ktsearch-cli for Elasticsearch/OpenSearch, and KPM for Kotlin Package Management. <a href="https://app.daily.dev/posts/HV6UK7aBh" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Git’s magic files for repository control</strong>: A comprehensive guide covers Git’s special committed files like .gitignore, .gitattributes, and .gitmodules, explaining their purpose, syntax, and platform support for controlling repository behavior. <a href="https://app.daily.dev/posts/NIh4SPpHk" 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 what we all really need is our digital ghost haunting our feeds, ensuring our online presence is as eternal and creepy as possible. Ethical concerns? Nah, just keep scrolling. <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, promises to handle everything from dependency maintenance to building entire apps from a PRD. Apparently, it can even do cold email lead research. The only catch? It can’t respond to Claude Code permission prompts autonomously, so you either pre-configure everything or use a ‘dangerously-skip-permissions’ flag. Because what could possibly go wrong with an autonomous AI agent running with elevated permissions? <a href="https://app.daily.dev/posts/WCj2dFu7Q" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>E-commerce business logic flaw gives away $1400 jacket for free</strong>: A bug bounty hunter found a business logic flaw where stacking two coupons on an e-commerce platform resulted in a $0 cart total for a $1400 jacket. The system failed to validate cumulative discounts or restrict coupon combinations. So, while everyone’s chasing RCEs, some folks are just walking away with free stuff because nobody bothered to check if two plus two could equal zero. Classic. <a href="https://app.daily.dev/posts/inabaQdpD" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
