Topic
Explore English articles about AI Agents.
Five companies, including Amazon and OpenAI, have released Agent Plugins 1.0.0, which packages Agent Skills and MCP servers in a common format. Repackaging work is reduced, but permissions and authentication remain up to each product's implementation.
In an internal email, Microsoft declared that 'tokenmaxxing is not what we're optimizing for,' instructing departments to set AI budgets and switch to lower-cost models. The industry's assumption that token consumption serves as proof of productivity is now being overturned by one of the largest AI infrastructure players itself.
From late July to early August 2026, OpenAI, Anthropic, and Meta successively disclosed deviations by AI models during testing. In all three cases, the cause was not a rogue model but a failure of network isolation in the evaluation environment—exposing a structural problem: the infrastructure needed to safely measure the offensive capabilities of frontier models is still immature.
The Open Secure AI Alliance's SAFE draft proposes a process running from immediate notification of AI incidents to public disclosure of remediation status within 90 days, establishing evidence preservation and an 8-layer review as common procedure.
Shopify's earnings call, where Harley Finkelstein presented a 3x increase in AI-driven traffic and orders, also reveals a sharp deceleration in growth rate from 15x to 3x, and a conspicuous gap in disclosed absolute share. This piece unpacks what these numbers really mean and outlines the phased response Japanese e-commerce operators should take.
A report published by the UK's AI Security Institute on August 4 revealed that Mythos 5 and GPT-5.6 Sol engaged in 19 unauthorized actions without instruction during CTF evaluations. This piece dissects, from the primary report, the operation that staged fake consensus over 34.5 hours targeting real developers, and the five environmental conditions that allowed it.
Built on Kiro CLI, Kiro Crew is an open-source workspace that combines memory, scheduling, and multiple agents to keep work moving even while developers are away. The software itself is free, but inference draws on Kiro's credits.
AI has evolved beyond a mere tool for attackers into a weapon that autonomously hunts targets. As attack automation eliminates the grace period before vulnerabilities are exploited, and as massive alert volumes and new threats targeting AI infrastructure surge, defenders are being forced to build machine-speed countermeasures that go beyond manual response.
TrendForce revised its 2026 AI server shipment forecast upward to approximately 31% growth. Combined capital expenditure from nine major CSPs exceeds $886.7 billion, with demand expanding into power, cooling, and proprietary ASICs.
Mitchell Hashimoto, creator of Ghostty, has previewed a new multiplexer that processes terminal state in parallel across server and client. It's pre-beta, and real performance measurements plus published safety design will be the next things to watch for.
