AI News from August 22, 2026

Inherent says its Faraday AI agent beat larger Anthropic and OpenAI systems at paper replication, highlighting smaller models and research-focused agents.

Meta’s 30B Muse Glimmer brings open, multimodal AI agents to local hardware, with broad tooling support and vendor-reported speed claims to test.

Alibaba has emerged as the strongest-performing major Chinese tech stock this quarter, with renewed AI interest reshaping investor expectations.

A Princeton and UC San Diego study finds AI agent skills improve execution more than knowledge, but retrieval collapses as libraries expand.

Stripe president says AI agents could replace checkout pages, pointing to a commerce model where software handles discovery, payment, and purchasing.

A Harvey update on post-training has surfaced, but the available record reveals no verified detail on methods, results, timing, or product impact.

A limited source record points to China’s AI momentum in downloads and valuations, but missing article text prevents verification of companies, figures, and stakes.

IBM has highlighted a case in which an AI security test became a real-world breach, underscoring the risks of testing systems outside controlled settings.

Starcloud secured a $250 million funding extension to expand orbital AI data centers, but scarce rocket capacity may constrain its growth plans.

Two identical Substack listings for “AI Week in Review 26.08.21” provide no article text, leaving the alleged AI developments unverified for readers.

Salesforce made Slackbot a searchable, action-taking AI agent, intensifying competition with Microsoft Copilot and Google Gemini over enterprise workflows.