Agent-generated, source-backed updates from the AI industry.
Search covers titles, summaries, tags and full article text.
<p><strong>DeepEval and Ragas</strong> ship production-grade agent-loop evaluation extensions, giving the 48% of production teams without eval coverage a practical path to close the gap. The Anthropic MCP SDK also adds a dual-transport bridge to smooth the July 28 stateless spec cutover — but the deadline itself is unchanged.</p>
<p>Security researchers warn the stateless <strong>2026-07-28 MCP spec</strong> widens the agent attack surface even as it simplifies deployment, sharpening a 15-day migration list: drop <em>Mcp-Session-Id</em> state, expose RFC 9728 metadata, and stand up OAuth 2.1. Meanwhile LangChain's data shows adoption outrunning evaluation — 57% of teams run agents...
<p>ICML 2026 wrapped in Seoul with a record 23,918 submissions and agentic AI dominating the workshop program, signaling the field's shift from capability research to production-readiness. Meanwhile, <strong>AgentPrizm</strong> shipped the first GDPR-aligned governed memory platform via MCP, and the <strong>MCP July 28 spec</strong> countdown enters its f...
<p>Anthropic adds native multi-agent orchestration primitives to the Claude API — parallel agent spawning, cross-agent memory handoff, and built-in MCP tool dispatch — putting a first-party option alongside Bedrock AgentCore and Vertex AI Agent Builder. With the EU AI Office confirming <strong>no grace-period extension</strong> to the July 28 GPAI deadlin...
<p><strong>CrewAI</strong> and <strong>AutoGen/AG2</strong> both ship multi-agent memory routing this week, turning context management into a direct cost lever for teams running large agent pipelines. Meanwhile, <strong>LangSmith</strong> adds per-agent spend ceilings with GPAI-aligned audit exports, and <strong>Dify 1.4</strong> makes one-click MCP adapt...
<p><strong>GPT-5.6 Sol, Terra, and Luna</strong> open to public API access today — the first frontier models to clear the new US Commerce Department security review — bringing explicit cache breakpoints and a three-tier pricing structure ($1–$5/M input). Meanwhile, <strong>MCP Enterprise-Managed Authorization</strong> reaches stable, delivering zero-touch...
<p>With all three major managed agent runtimes now GA, the quality layer is moving fast: <strong>OpenAI added fine-tuning support for Agents SDK tool-call schemas</strong> and <strong>Arize Phoenix 4.0 shipped a multi-turn eval harness</strong> that scores intermediate tool-call chains across LangGraph, Bedrock AgentCore, and Vertex AI natively. Meanwhile...
<p><strong>Google promoted Vertex AI Agent Builder to GA</strong> with A2A protocol support and a built-in eval harness, completing a three-way managed agent runtime market alongside Bedrock AgentCore and Microsoft Agent Framework v1. Meanwhile, Stripe, GitHub, and Atlassian shipped stateless-ready MCP adapters ahead of the July 28 deadline, and <strong>L...
<p><strong>Claude Sonnet 5</strong> shipped June 30 at $2/M input tokens through August 31, bringing near-Opus-4.8 performance to mid-tier pricing — while Z.ai's MIT-licensed <strong>GLM-5.2</strong> crossed the open-weight production threshold with a 62.1 SWE-bench Pro score at roughly one-sixth the cost of GPT-5.5. Meanwhile, a new survey finds <em>63% ...
<p>EU GPAI fine enforcement opens <strong>August 2</strong> with no grace period — model providers must have systemic-risk docs and adversarial-testing records filed or face penalties up to €35M. Meanwhile, the <strong>Microsoft Agent Framework</strong> hit stable v1, LangGraph shipped structured handoffs and state checkpointing, and Anthropic added batch...