<h2>Must know</h2>
<ul>
<li><strong>OpenAI undercuts Google with GPT-5.4 Nano and Mini</strong>: OpenAI has launched GPT-5.4 Mini and Nano, targeting high-volume subagent tasks with a 400k context window. The Nano model is priced at a mere $0.20 per million input tokens, specifically positioned to beat Google’s Gemini 3.1 Flash-Lite on cost. These models are already rolling out to GitHub Copilot, emphasizing a multi-model architecture where small models execute parallel subtasks planned by larger ones. <a href="https://app.daily.dev/posts/6TGfnHW2k" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Firefox 149 adds built-in VPN and Sanitizer API</strong>: Launching March 24, Firefox 149 will include a free built-in VPN with a 50GB monthly data cap for users in the US, UK, Germany, and France. More importantly for security, Firefox 148 and 149 are the first to implement the Sanitizer API, a new web standard designed to natively neutralize XSS attacks. The update also introduces productivity features like Split View and Tab Notes to combat its declining market share. <a href="https://app.daily.dev/posts/nPBbJz5BW" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Mistral Forge challenges OpenAI in custom enterprise training</strong>: Mistral announced Forge at Nvidia GTC, a platform allowing enterprises to train custom models from scratch on proprietary data rather than just fine-tuning. Unlike RAG-heavy competitors, Forge supports full pre-training and reinforcement learning for agentic systems, targeting high-compliance sectors like finance and government. Early partners include ASML and the European Space Agency, signaling a shift toward deep model ownership for large-scale infrastructure. <a href="https://app.daily.dev/posts/6pejPwPmX" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Vite 8 is officially out featuring Rolldown for 10x faster builds</strong>: The major release replaces Rollup with the Rust-based Rolldown, significantly improving build performance and dev server start times. <a href="https://app.daily.dev/posts/db4oA9qt6" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>CloudNativePG brings self-healing PostgreSQL to Kubernetes</strong>: The open-source operator uses a shared-nothing architecture and declarative management to handle automated failover and scaling in K8s environments. <a href="https://app.daily.dev/posts/Ae6hH5fZK" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>LangSmith Sandboxes provide isolated microVMs for untrusted agent code</strong>: A new private preview feature offers ephemeral, Docker-compatible environments to safely execute code generated by AI agents. <a href="https://app.daily.dev/posts/Uz5xAKtF3" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Oracle revives Project Detroit for Java and Python interoperability</strong>: The project embeds V8 and CPython runtimes directly into the JVM process using the Foreign Function and Memory API for high-performance scripting. <a href="https://app.daily.dev/posts/aQEMjwL5W" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NVIDIA CloudXR 6.0 brings RTX streaming to Apple Vision Pro</strong>: Native visionOS integration allows streaming high-fidelity 3D applications from workstations or the cloud with dynamic foveated rendering. <a href="https://app.daily.dev/posts/Lajg4CAxX" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GSD framework aims to solve context rot in Claude Code</strong>: This lightweight meta-prompting system uses structured XML plans and atomic commits to maintain AI output quality as context windows fill up. <a href="https://app.daily.dev/posts/NtBuG2cMQ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Bun v1.3.11 adds native OS cron job scheduling</strong>: The update introduces Bun.cron for cross-platform task scheduling alongside 105 bug fixes and improved Windows ARM64 support. <a href="https://app.daily.dev/posts/DuG54Q8tB" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>World ID wants to be the power of attorney for your AI</strong>: Tools for Humanity (formerly Worldcoin) is launching AgentKit, a way to link your iris-scanned identity to AI shopping agents so they can spend your money. It’s a desperate bid for relevance for a token that’s down 76%, framed as ‘agentic commerce’ when it’s really just a digital leash for bots. If you’ve ever wanted to cryptographically delegate your financial mistakes to a hallucinating LLM, this is for you. <a href="https://app.daily.dev/posts/63OP7gZFo" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The $12.5 million band-aid for AI-generated bug slop</strong>: Big Tech is throwing a few million at the Linux Foundation to help maintainers deal with the absolute deluge of garbage security reports their own AI tools are creating. Instead of making the tools less prone to hallucinating vulnerabilities, they’re funding ‘triage strategies’ to help humans clean up the mess. It’s the tech equivalent of a company selling you a leaf blower and then charging you for a rake to clean up the neighbor’s yard. <a href="https://app.daily.dev/posts/LS9KJQCC0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Krafton CEO uses ChatGPT to dodge executive bonuses</strong>: In a peak ‘AI-enabled management’ move, Krafton’s CEO reportedly consulted ChatGPT for ways to fire executives and avoid paying out Subnautica 2 earnout bonuses. A Delaware judge wasn’t impressed by the prompt-engineered legal strategy and ordered the company to rehire the director. It turns out ‘How do I commit breach of contract?’ isn’t a query that holds up in a court of law, even if the LLM gives you a bulleted list. <a href="https://app.daily.dev/posts/wVjQNYbQm" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
