Agent-first stacks: hybrid compute, autonomous ops, and software factories

DataAgent emerges from stealth with $10M pre-seed to build autonomous agents that fix cloud infrastructure failures. The startup launches agents that autonomously detect and repair cloud-production failures, demonstrating a concrete pattern for automated incident response that outcome engineers must integrate with observability, rollback, and audit capabilities — think Principle 14 in practice.

PRs NOT Welcome: How Top AI Open Source Projects Are Managing Thousands of Contributors. Swyx shows projects running internal agent-driven software factories to triage, author, and land PRs, giving a production blueprint for agent orchestration and contributor governance — a practical reference for Principle 03 and Principle 09.

Introducing agentic video understanding with Gemini. DeepMind adds agentic video understanding that cuts token use up to 88%, enabling cheaper, perceptual agents and tighter context engineering for multimodal workflows; outcome engineers should plan for video-aware agent loops and new validation needs (Principle 09 and Principle 16).

Perplexity launches Hybrid Compute to split workloads between frontier cloud models and local LLMs for Mac users. Perplexity routes sensitive and cost-sensitive requests to local LLMs while offloading heavy reasoning to cloud frontiers, modeling a hybrid architecture that outcome teams can adopt to balance privacy, latency, and cost (Principle 07 and Principle 15).

Wafer raises $40M Series A at $200M+ valuation to deploy agents optimizing open-source models. Wafer funds agentic infrastructure that optimizes model choice and inference for business workloads, highlighting operational tooling you’ll need for efficient, observable inference pipelines and decisioning (Principle 12 and Principle 09).