<h2>Must know</h2>
<ul>
<li><strong>OpenAI launches GPT-5.4 mini and nano for subagent orchestration</strong>: OpenAI has released GPT-5.4 mini and nano, targeting high-volume subagent tasks with a 400k context window and significantly lower pricing than Google’s Gemini 3.1 Flash-Lite. The nano model is optimized for classification and data extraction, while the mini variant is rolling out to GitHub Copilot with a 0.33x request multiplier. This release reinforces the industry shift toward multi-model architectures where larger models plan and smaller, faster models execute parallel subtasks. <a href="https://app.daily.dev/posts/6TGfnHW2k" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Mistral Forge and Leanstral target enterprise code verification</strong>: Mistral announced Forge, a platform for training custom models from scratch on proprietary data, alongside Leanstral, a specialized agent for formal code verification using the Lean language. Leanstral aims to solve the reliability gap in AI-generated code by grounding output in formal proofs and tests, claiming significantly lower costs than Anthropic’s Claude Sonnet for similar verification tasks. These tools are designed for “agent-first” environments where autonomous systems manage their own training and compliance pipelines. <a href="https://app.daily.dev/feed-by-ids?id=XdxpqgLtJ&amp;id=6pejPwPmX" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NVIDIA CloudXR 6.0 brings RTX streaming to visionOS</strong>: NVIDIA and Apple have integrated CloudXR 6.0 with visionOS, allowing the Apple Vision Pro to stream high-fidelity, RTX-powered 3D applications with dynamic foveated rendering. This native Swift framework enables enterprise developers to run complex digital twins and simulations from workstations or the cloud directly to the headset. Partners like BMW and Volvo are already using the integration for automotive design reviews and factory visualization. <a href="https://app.daily.dev/posts/Lajg4CAxX" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Uber Michelangelo evolves to Kubernetes mesh with 100 custom CRDs</strong>: Uber bypassed etcd scaling limits using a MySQL-backed storage abstraction to handle 30 million predictions per second across a multi-cloud mesh. <a href="https://app.daily.dev/posts/rfWF0zX0x" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>ConnectRPC enables typed SDK generation from GraphQL operations</strong>: The new protocol layer in Cosmo Router compiles GraphQL into Protobuf, serving gRPC and typed Go/TS clients with backward compatibility locks. <a href="https://app.daily.dev/posts/Cgx6Z8AyK" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>LangSmith Sandboxes provide microVM isolation for untrusted agent code</strong>: A new private preview offers ephemeral, Docker-compatible environments to safely execute AI-generated code with native tracing and autoscaling. <a href="https://app.daily.dev/posts/Uz5xAKtF3" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Bun v1.3.11 adds native cron scheduling and ANSI-aware slicing</strong>: The update includes Bun.cron for OS-level task scheduling and fixes critical bugs in HTTP/2 and Windows ARM64 shims. <a href="https://app.daily.dev/posts/DuG54Q8tB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Vite 8 integrates Rolldown for 30x faster production builds</strong>: The major release focuses on performance parity between dev and build modes by leveraging the Rust-based Rolldown bundler. <a href="https://app.daily.dev/posts/CZHeTFzvD" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Zeroboot achieves sub-millisecond microVM forking for AI agents</strong>: The open-source project uses Firecracker and copy-on-write forking to spin up isolated agent sandboxes in approximately 0.8ms. <a href="https://app.daily.dev/posts/6LGcsCCdV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>PostgreSQL 18 deprecates MD5 in favor of SCRAM-SHA-256</strong>: The upcoming release issues warnings for MD5-encrypted passwords as a precursor to full removal in future versions. <a href="https://app.daily.dev/posts/nekgWU5A0" 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 shopping bot</strong>: Tools for Humanity launched AgentKit, a way to link your iris-scanned identity to AI agents so they can spend your money. It’s a bold pivot for a project whose token is down 76%, essentially betting that the solution to ‘AI bot spam’ is more crypto-biometric gatekeeping. If you’ve ever wanted to cryptographically delegate your financial ruin to a LLM, the Orb is waiting. <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 funding a Linux Foundation initiative to help maintainers triage the absolute deluge of garbage security reports generated by the very AI tools these companies sell. We’ve reached peak circular economy: sell the ‘productivity’ tool that breaks open source, then donate a fraction of the profits to build a filter for the mess. It’s not a security initiative; it’s a janitorial service for the LLM era. <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>: A Delaware judge ordered the rehiring of the Subnautica director after Krafton’s CEO literally asked ChatGPT for ways to avoid paying out earnout bonuses. The AI apparently suggested firing the execs right before their revenue targets hit—targets Krafton itself made impossible by delaying the game. It’s a perfect case study in ‘AI-driven leadership’ being used as a thin veil for basic contract violation. <a href="https://app.daily.dev/posts/wVjQNYbQm" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
