Li Lin-backed crypto-friendly securities platform UMX officially launches.

Upbit official announcement: Due to BitMEX terminating access to the travel rule solution VerifyVASP, Upbit will suspend digital asset deposits and withdrawals of 1 million KRW or more between the two exchanges starting October 16. Deposits under 1 million KRW will remain supported as normal. Manual deposits of 1 million KRW or more from BitMEX to Upbit will be discontinued from 14:00 on December 18. Any such assets received after that will enter a refund process, and users must retrieve them in accordance with the refund procedure.

Trader BonkGuy’s portfolio value has surged by over $1 million in the past 24 hours, now standing at $11.43 million. His current positions boast unrealized gains of roughly $5.63 million from PONS, $1.57 million from USELESS, and $1.43 million from MarsCoin, keeping him the top-ranked trader on the Fomo platform.

According to EmberCN’s monitoring, since the Papertrade platform on the Hyperliquid chain launched loss mining on the 10th, the three addresses with the highest trading volumes have all surpassed $10 billion each in less than six days. The top address posted a trading volume of $11.6 billion, with trading losses of $780,000, earning 46 million PAPER tokens at an acquisition cost of $0.017 per token. The second address recorded $10.3 billion in trading volume, $930,000 in trading losses, and received 85 million PAPER tokens at an acquisition cost of $0.011 per token. The third address saw $10.1 billion in trading volume, $980,000 in trading losses, and obtained 252 million PAPER tokens at an acquisition cost of $0.0039 per token. The acquisition cost excludes PAPER token dividends; factoring in dividends would lower the cost further, with the third address potentially even holding a negative cost.

As the market rebounds, Bonk Guy(@theunipcs)'s portfolio has grown by $1M+ in the past 24 hours. He remains No. 1 on both #FOMO's 24H and All-Time leaderboards.

Hong Kong Exchanges and Clearing (HKEX) has published a consultation summary document on shortening the settlement cycle of Hong Kong’s spot stock market, which shows widespread market support for implementing a T+1 settlement cycle in Hong Kong’s securities market. Based on feedback from various stakeholders, HKEX plans to complete the relevant transition in the second half of 2028, provided the market is ready and regulatory approval is obtained. The consultation received a total of 109 non-duplicate responses from market stakeholders including brokers, custodians, investment managers, industry associations, technology vendors, and investors. Respondents supported Hong Kong’s adoption of a T+1 settlement cycle, agreeing that this move will help enhance market efficiency, reduce settlement risks, improve capital efficiency, and consolidate Hong Kong’s competitiveness as an international financial center.

Beating AI News: Microsoft announces completion of pre-training for its new self-developed model MAI-Base-2. The model has a total of 2.8 trillion parameters, adopts a Mixture-of-Experts (MoE) architecture, and activates only 80 billion parameters per forward pass. Between August and September this year, Microsoft used 16,000 NVIDIA GB200 GPUs for 38 days of training, processing approximately 50 trillion tokens. The new model also introduces native visual pre-training for the first time, with visual data accounting for 18% of its training data. Internal tests released by Microsoft show that MAI-Base-2 outperforms models including DeepSeek-V4-Pro-Base, Kimi K3, and Qwen3.8 on text prediction benchmarks for code, question answering, mathematics, and science and engineering disciplines. Microsoft also claims that training a model of equivalent performance to DeepSeek-V4-Pro-Base would require only 1/4 to 1/12 of the computing power using its approach. However, Microsoft has invested additional computing power to build a larger, more capable model; MAI-Base-2’s actual training computing power consumption is still higher than that of DeepSeek. Microsoft AI head Mustafa Suleyman stated at an internal meeting on October 15 that MAI-Base-2’s pre-training base is already stronger than the OpenAI model supporting GPT-5.6 Sol, while requiring only one-fourth the training computing power. Microsoft improved MoE expert scheduling, Muon optimizer, and data filtering methods to reduce GPU memory waste and enhance the stability of large-scale training. The company also emphasized that all training data is processed in-house, and the model’s capabilities were not obtained through distillation of other firms’ models. MAI-Base-2 is currently in the post-training phase. Microsoft plans to first deploy it for its own products, then license the model weights to enterprise partners via Foundry.