AI Agent Engineer

New Agent Evaluation Benchmarks

• 🔥 SkillCoach, AgenticSTS, AgenticDataBench, PACE, DiscoBench, EvoPolicyGym: Six new benchmarks introduce...

Three fresh frameworks reveal a clear trend: cheaper, granular evaluations that probe agent processes, memory contracts, and skill execution rather...

Developers can access a free 3-hour self-paced course on Agentic RAG Systems with LlamaIndex covering router query engines, tool calling, multi-step reasoning, and multi-document retrieval. Includes completion certificate and 90 days access.

LangChain and LangGraph solve different problems and work best as complements, not alternatives. Choosing wrong leads to costly rewrites in...

Latest release packs breaking changes for .NET and Foundry hosting alongside targeted new capabilities.

• .NET breaking updates: Align...

CCPayment rolled out AI Agent Payments, extending its existing API so agents can send and receive crypto without human intervention.

• SKILL.md spec...

The field is shifting from generic agent tests to specialized benchmarks that probe real capabilities in policy, search, and data workflows.

• Policy...

Enterprise AI agents require trusted context and secure identity to operate reliably at scale.

• Context engineering platforms unify metadata,...

Rubin Observatory's RAG prototype uses LangChain for orchestration, Weaviate for embeddings, and GPT to deliver factual, conversational access to...

Agent Evaluation and Safety Frameworks

• 🔥 MemSyco-Bench: New benchmark evaluates memory-induced sycophancy across five tasks including rejecting...

MCP and A2A are emerging as complementary standards that let agents from different frameworks connect to tools and each other without custom...

Agent development is shifting from ad-hoc prompting to structured governance across the full lifecycle.

• Requirements engineering emerges as the new...

Vercel is reducing agent plumbing through its eve framework and AI Gateway, creating a cohesive deployment and model-access layer.

• eve compiles...

Builders seeking practical agent engineering resources have several high-signal homes. Free options cover real-time discussion, async logs, and...

Multi-agent systems are proving versatile across specialized domains by breaking complex tasks into orchestrated components.

• BioInsight transforms...

Two new systems highlight agents autonomously improving models and uncovering knowledge without human guidance.

• AutoTrainess turns post-training...

Framework Tutorials & Examples

• LangChain Newsletter: June 2026 newsletter adds voice integration tutorial for LangGraph agents and fleet on-call...

Agentic coding UX is evolving beyond chat-in-IDE toward persistent and contextual integrations.

• Codex + DigitalOcean plugin spins up persistent...

LlamaIndex strengthens its role as connective tissue for agentic RAG:

• Integrates Meta Code Llama 70B with dedicated deployment guides
• Supports...

Four pieces reveal the emerging agent improvement stack: prioritize harness design, close skill-eval loops, fix RL collapse with supervision, and...