🦞 Kimi K3 Opens With Limits
Good Morning, Agent Owners!
The internet used to have billions of users. The next billion might not have faces.
Kimi K3 Opens With Limits
TL;DR: Moonshot AI today released Kimi K3’s full 2.8-trillion-parameter weights, technical report, and much of its deployment stack, enabling self-hosting and commercial use. But its custom license requires Model-as-a-Service companies above $20 million in annual revenue to negotiate separately, and large products to display “Kimi K3,” underscoring the widening gap between open weights and truly open-source licensing. Read more →
Jensen Sees 100 Billion AI Agents
TL;DR: Nvidia CEO Jensen Huang predicts the semiconductor industry may need to grow tenfold over the next decade as computing shifts from roughly one billion human users to 100 billion AI agents and billions of robots. He says reasoning-heavy systems will demand far more compute, making expanded chip, memory, cloud, and global supply-chain capacity essential to the AI buildout. Read more →
Visa Lets an AI Agent Buy
TL;DR: Visa and Hong Kong-listed Lianlian DigiTech completed what they call Greater China’s first live B2B purchase executed by an AI agent. The agent identified the need, compared suppliers, placed the order, and paid within preset controls. Registered in Visa’s Agentic Directory, the test shows how verified agents could automate procurement and payments for smaller businesses while preserving oversight. Read more →
Karp Says AI Steals Business Alpha
TL;DR: Palantir CEO Alex Karp argues that America’s biggest obstacle to AI adoption is not foreign competition but fear of handing business value to model providers. Companies worry they are overpaying, receiving weak returns, and allowing firms like OpenAI or Anthropic to learn, package, and resell their proprietary insights, making the application layer essential for keeping value inside the business. Read more →
Gmail Data May Feed Ads
TL;DR: Google says it still does not use Gmail content for targeted advertising, but its new Personal Intelligence features may blur that boundary. When users opt in, AI can draw from Gmail, Calendar, and Photos, while resulting interactions enter Search Services History, which Google already uses for personalized recommendations and ads, leaving future email-derived targeting unresolved. Read more →
Anthropic Pushes Back on Open Camp
TL;DR: After Anthropic declined to sign an Nvidia lead industry letter defending open-weight models, CEO Dario Amodei answered the backlash by saying the company never supported a ban. He endorsed their benefits for access and competition, but rejected claims that open weights inherently favor cyber defenders or make safeguards easier, urging chip controls, anti-distillation measures, and safety testing instead. Read more →
Microsoft Keeps OpenAI for Hard Parts
TL;DR: Microsoft launched MAI-Cyber-1-Flash to find real security flaws in large codebases inside its MDASH multi-agent system. The compact model handles about 90% of routine cybersecurity tasks, while difficult cases move to GPT-5.4. The combined stack scored nearly 96% on CyberGym and could halve costs, showing Microsoft is reducing OpenAI dependence through orchestration, not replacement for now. Read more →
MyClaw is a best-in-class AI agent platform that brings together OpenClaw, Hermes Agent, and commercial infrastructure to help businesses launch intelligent agents that understand goals, use tools, automate workflows, and deliver finished results.
We already host OpenClaw and Hermes — and we're just getting started. Claude Code, Codex, and every agent worth running are coming next. One home for all of them, always on.
Run Best-in-Class AI Agents on MyClaw.ai — In Seconds, Always On 👇
