Agentic AI, LLM engineering, and the models behind modern automation — multi-agent systems, LLM evaluation and comparisons, RAG, fine-tuning, AI infrastructure, security, and production AI engineering.
From the blog
deepagents v0.5 ships harness profiles, async subagents, and Anthropic prompt caching baked in. We unpack when this opinionated harness beats raw LangGraph for production agents.
Where Claude agents, skills, and MCP are heading — longer-horizon autonomy and agent ecosystems — plus concrete moves to prepare your team and code.
The metrics that prove a Claude agent works — task completion, eval pass rate, escalation reasons, and cost per resolved task — plus what to ignore.
A realistic problem-to-shipped Claude agent walkthrough — scoping the task, narrow tools, evals from real tickets, and a staged production rollout.
Failure scenarios and concrete containment for production Claude agents — tool scoping, approval gates, budgets, kill switches, and eval gates.
The concrete skills and role shifts teams need so Claude agents and skills work across an organization — upskill vs. hire, and the roles that emerge.
Grow Claude Agent Skills from one team to many without sprawl or chaos — federation, a discovery layer, promotion tiers, and continuous curation.
Honest trade-offs for Claude Agent Skills: where they win, where a prompt or script beats them, and which tasks you shouldn't automate yet.
The governance, trust, and safety guardrails leadership needs before scaling Claude Agent Skills — least privilege, risk tiers, logging, and review.
