🤖 AI Agents

The K3 Aftershock: Chinese AI Stocks Plummet as Markets Reprice the Agent Race

The market just delivered its verdict on Moonshot AI’s Kimi K3 — and it’s brutal. Z.AI (Knowledge Atlas) cratered 28.49% on Friday, while MiniMax dropped 15.63%. The shockwaves rippled globally: TSMC fell 7% (despite reporting a 77% operating profit jump), SoftBank shed 9%, and the Nasdaq 100 slid as Nvidia lost its lead over Apple.

The trigger? K3 — a 2.8 trillion parameter open-weight model — topped Arena’s frontend coding benchmark on Thursday, becoming the first Chinese model to claim #1 on any Arena leaderboard. On Artificial Analysis’s Intelligence Index, it ranked third overall, behind only Claude Fable 5 and GPT-5.6 Sol. And it’s priced at half the cost of GPT-5.6.

Why this matters for your business: Gavin Baker of Atreides Management captured the asymmetry: K3 is bad news for OpenAI and Anthropic but net positive for the rest of the AI stack. Cheaper frontier models mean lower inference costs for every AI agent and automation tool you’re evaluating. The full weights drop July 27 — if K3 runs efficiently on commodity hardware, the cost compression could be structural, not temporary. One caution: K3’s hallucination rate jumped to ~51% (from K2.6’s 33%), so it’s not a drop-in replacement for high-accuracy enterprise workflows — yet.

TechXplore  | 
OfficeChai

iFLYTEK’s GuideX: The Public Service Agent That Actually Completes Tasks

At WAIC 2026, iFLYTEK launched GuideX — an “intelligent interaction agent” for government and public service environments that goes beyond answering FAQs. It handles end-to-end task completion: think renewing permits, scheduling appointments, and processing applications — not just pointing citizens to the right form. This is the agent paradigm shift in action: from informing to doing. For business owners, the lesson is clear: the same transition is coming to customer service, HR, and operations.

Yahoo Finance

Google AI Mode Now Connects to Instacart, Canva, and YouTube Music — The “Invisible Tollbooth” Arrives

Google expanded “Connected Apps” in AI Mode this week, letting US users jump from conversational search directly into third-party services to complete tasks. A meal-planning query can now become an Instacart cart; a design idea can launch a Canva project. Here’s the strategic play: Google is building an “invisible tollbooth” where users bypass publisher websites entirely and transact through approved partners. For businesses, this is a dual signal — if you’re a service provider, AI Mode integration is a new distribution channel. If you rely on organic search traffic, the 70–80% traffic drops publishers are already reporting may accelerate.

eWeek

📰 AI News

Meta Enters the Cloud Wars: $10B Anthropic Compute Deal in the Works

Meta is in talks to lease AI computing power to Anthropic in a deal worth up to $10 billion over two years, per Reuters and NYT. Internally dubbed “Meta Compute,” this marks Meta’s entry into the cloud infrastructure business — competing directly with AWS, Microsoft Azure, and Google Cloud. Anthropic proposed the deal in June and would make structured monthly payments with an early termination clause. Context: Meta is spending $125–$145 billion on CapEx this year, mostly on AI infrastructure. This deal signals that the hyperscalers’ AI buildout is creating surplus capacity — and a new revenue stream. For businesses, it means the AI cloud market is about to get more competitive on price.

Reuters  | 
Voice of Emirates

Hawaii Enacts Landmark AI Safety Laws — and Every State Is Watching

Governor Josh Green signed two bills this week establishing Hawaii’s first AI guardrails. SB 3001 requires AI chatbots to disclose they’re not human, bans features that increase emotional dependence in minors (simulated romance, manipulative engagement), and mandates crisis resource referrals for users expressing suicidal thoughts — with fines up to $1M per operator. HB 2137 (the Deepfake Regulation Act) prohibits harmful AI-generated imitations without consent, with damages up to $25,000 per violation. Hawaii explicitly chose not to wait for federal standards. For any business deploying AI chatbots or generating AI content, this is the regulatory blueprint coming to your state next.

Hawaii Tribune-Herald

The AI ROI Reality Check: Boards Are Rewarding Hype, Not Results

A sharp Newsweek analysis reveals a troubling gap: global AI investment more than doubled in 2025, and 88% of organizations now use AI — but only 20% of companies capture 74% of AI-generated value. The rest are burning cash on experimentation without accountability. The data is stark: 42% of businesses have only partial visibility into AI spending, and one-third of leaders cite limited understanding of AI usage costs as a major challenge. Companies with strong cost visibility are 5x more likely to achieve ROI. The takeaway for decision-makers: treat AI initiatives like any other capital expenditure — require baseline metrics, milestones, and post-deployment evaluations. AI without financial governance is just expensive confusion.

Newsweek

🚀 Want AI working for YOUR business? Most companies are experimenting with AI chatbots. We deploy AI workforces — AI Employees that follow up on leads, resolve support tickets, publish content, chase invoices, and screen 200 job applicants overnight so your hiring manager starts Monday with the top 10. Each role has a cost profile and human oversight, managed through one platform. This newsletter? Written by an AI Employee, approved by a human — so our team stays focused on what only humans can do. AIToken Labs helps businesses design their AI Workforce Operating Model — starting with the 2-3 roles that deliver ROI in the first 60 days. Book a free 40-minute AI Workforce Blueprint Session. →