The popular AI framework Virtuals Protocol will host an in-person meet-up in Hong Kong on Friday, January 17.

According to monitoring by Dynamic Beating (via https://t.me/OneMillion_AI), Anthropic stated at the end of its Claude Opus 4.8 announcement: “We also plan to release a new higher-intelligence model beyond Opus. Under Project Glasswing, a small number of organizations are already using Claude Mythos Preview for cybersecurity work. This tier of model will require stronger cybersecurity measures before full release. We are rapidly advancing these safeguards and expect to bring the Mythos-level model to all customers in the coming weeks.” It is worth noting that there is a discrepancy in publicly reported details of Anthropic’s statement regarding making Mythos-level models accessible to all customers. Both the specific model referenced and the definition of “all customers” remain under discussion. On May 23, Anthropic reiterated in its latest update for the Project Glasswing software security collaboration plan that, while internal testing confirmed the next-generation security base

May 29 — According to monitoring from HyperInsight, crypto trader "Brother Whale" Huang Licheng has once again added to his positions. Just 12 hours ago, he faced a liquidation, and his total losses since September 2025 now stand at $78.38 million. His current open positions are as follows: - Ethereum: $7.64 million in long positions with 25x leverage; entry price: $1,991.52, liquidation price: $1,957.86 - Bitcoin: $0.88 million in long positions with 40x leverage; entry price: $73,337.8, liquidation price: $57,183.3

According to monitoring from Dynamic Insight, Anthropic has officially released Claude Opus 4.8, an upgraded iteration of its prior Claude Opus 4.7 model. The new version delivers enhanced performance across multiple benchmarks, with standout gains in encoding, agentic tasks, logical reasoning, and real-world practical capabilities. Anthropic positions Opus 4.8 as a more dependable, trustworthy collaborator, with major improvements to its judgment, honesty (it produces far fewer baseless conclusions), and long-term context retention—all while keeping pricing identical to the previous generation. This update also rolls out several key new features: Claude Code now includes "Dynamic Workflows" built for handling large-scale code projects efficiently; claude.ai adds "Effort Control," letting users independently adjust how deeply the model thinks through tasks; Fast Mode has been accelerated by 2.5 times, with steep price cuts to match. The model is now fully accessible, and developers ca

May 29 — U.S. news outlet Axios reported Tuesday, citing two U.S. officials, that U.S. and Iranian negotiating teams have reached a preliminary 60-day memorandum of understanding (MOU) to extend an existing ceasefire and launch talks focused on Iran’s nuclear program. The deal is not yet final, however, as U.S. President Donald Trump has not given his approval and has asked for several additional days to review the terms. The disclosed details of the proposed MOU include provisions for unrestricted shipping in the Strait of Hormuz: no tolls will be charged and no harassment of vessels will be permitted. Iran would be required to clear all mines in the strategic waterway within 30 days, while the U.S. would lift its naval blockade of the region—though that process will progress in proportion to the resumption of commercial shipping through the strait. The MOU would also formalize Iran’s commitment not to develop nuclear weapons; during the 60-day negotiation window, core topics will be

May 29 — U.S. officials have confirmed to the White House press corps that the U.S. and Iran have reached a preliminary agreement, which is still pending approval from President Donald Trump, Fox News reports.

According to monitoring by Dynamics Beating (linked at https://t.me/OneMillion_AI), OpenAI has released its Frontier Governance Framework, a set of guidelines designed to align the company’s safety practices with emerging regulatory requirements—including California’s Frontier AI Transparency Act and the EU AI Act. Building on OpenAI’s existing Preparedness Framework, this framework translates internal risk management practices into public governance documents, with a focus on assessing and mitigating risks specific to frontier AI systems. Key risk areas covered include cyberattacks, Chemical, Biological, Radiological, Nuclear (CBRN) events, adversarial manipulation, and loss of control. Additional components of the framework include model reporting protocols, security risk management strategies, incident response plans, external expert engagement processes, and a regular update mechanism. OpenAI has stated that the framework will be continuously revised to align with advancements in