SuperPlane secures $2.6M to turn production operations into an AI-native workflow layer. The open-source, AI-first control plane lets engineers and agents safely coordinate production operations and embed agentic workflows into existing systems. If you’re building agentic ops, this is a concrete orchestration layer to run, observe, and audit (Principle 09).

Haystack: Open-source AI Framework for Production-ready Agents and RAG. The framework wires retrieval, reasoning, and tools into a deployable stack so teams can move from prototypes to production RAG and agent pipelines. Use it to standardize context engineering and reduce bespoke glue when shipping outcome-driven agents.

Qwen-AgentWorld: Language World Models for General Agents. Alibaba releases language-based world models and AgentWorldBench to simulate multi-domain agent environments and evaluate simulation fidelity. That lets you run large-scale offline experiments and adversarial scenarios—directly useful for validating agent policies and auditing outcomes (Principle 16).

Seltz raises $12.5M to rebuild web search for AI agents. They’re building a search stack that returns machine-ready passages in milliseconds, optimized for agent retrieval needs. Lower latency and structured results change how you engineer context, reduce token waste, and improve agent decision quality.

Introducing computer use in Gemini 3.5 Flash. DeepMind adds built-in computer use so models can control apps, browse, and perform cross-platform tasks, shrinking the gap between instruction and action. That shifts agent design—expect more model-level tool interfaces and fewer external orchestrators, which requires new validation and safety patterns (Principles 07 & 09).