Anthropic2026-07-08 19:56:27Anthropic Restricts Claude Agent Access Amid AI Automation Boom in CryptoOn April 4, 2026, Anthropic cut Claude Pro/Max subscription coverage for third-party agent frameworks like OpenClaw, forcing crypto AI developers into pay-as-you-go billing. Single-day costs can reach $1,000–$5,000. Developers are migrating to OpenAI, local models, or Hermes Agent.460
Venice Token2026-07-08 13:54:14Erik Voorhees' Venice Token Faces Pump and Dump Allegations, Insiders Dumped $10.2MOnchain analyst Amir Ormu reveals that Venice project insiders sold $10.2 million worth of VVV tokens shortly after launch, causing an 63% price crash. Founder Erik Voorhees defends the transparency of tokenomics.630
B.AI2026-06-30 18:01:08B.AI Launches Team-Level AI Collaboration Product 'B.AI Team', Starts Grayscale Testing with Global RecruitmentB.AI officially announced the first phase of grayscale testing for its team-level AI collaboration product 'B.AI Team', and launched a global public recruitment for initial test teams. The product addresses enterprise AI collaboration pain points with a shared Credits pool for unified compute resource scheduling and cost visualization, plus multi-level permission settings for role-based model usage policies. Approved teams will gain priority access to the official version, experiencing a new paradigm of smart collaboration.450
AI training2026-06-29 00:31:36Dwarkesh Patel Proposes Next-Gen AI Training: Real-World Learning, OPSD, and "Dreaming" SimulationsDwarkesh Patel argues that the next generation of AI training should move beyond reinforcement learning from verifiable rewards (RLVR) toward continuous learning in real-world tasks, with experience written back into model weights. He highlights three key paths: malleability (the ability to continuously integrate new experience post-deployment), on-policy self-distillation (OPSD), and "dreaming" simulations. This shift from pre-training to post-deployment evolution could profoundly impact decentralized AI and on-chain agent development.540
Anthropic2026-06-28 22:01:21Anthropic CEO Warns US Congress: Open Source AI on 'Very Dangerous Path', Crypto AI Tokens Face New Regulatory RisksDario Amodei, co-founder and CEO of Anthropic, told US lawmakers on June 28 that open source AI is heading down a 'very dangerous path', as developers lose the ability to monitor abuse, revoke access, or update safety mechanisms once powerful models are released openly. Compared to closed-source systems, fully open models are harder to govern and may cause irreversible harm. The remarks sent ripples through crypto AI token markets, raising concerns that projects like Fetch.ai, SingularityNET, and Ocean Protocol could face stricter regulatory scrutiny. Amodei's stance highlights the tension between decentralized AI ideals and the need for controlled AI safety, prompting crypto projects to rethink open-source strategies.550
Anthropic2026-06-28 19:31:15Anthropic CEO Warns: Open-Source AI Enters 'Very Dangerous Path'; Control Loss Raises Alarm at U.S. CongressAnthropic 联合创始人兼 CEO Dario Amodei 在美国国会听证会上指出，开源 AI 模型正在进入“非常危险的路径”。一旦强大模型开源，开发者将无法监管滥用、撤销权限或动态更新防护机制，可能带来不可逆的滥用隐患。这一警告引发了对 AI 开放与安全平衡的深入讨论，也对加密领域去中心化 AI 治理提出了新挑战。540
