Cai Wensheng: AI transforms productivity, Web3 transforms production relations, and data will be the most valuable asset in the future.

At the "AI × New Finance – Global Innovation Tour Hong Kong Stop & Yangtze River Star Cluster Plan Launch Conference", Zhang Dayong, co-founder and CEO of Obita, stated that AI is compressing tasks such as contract drafting, coding, and business decision-making to seconds or even less, while traditional cross-border payments still take T+1 to T+7. This efficiency gap will drive a restructuring of payment infrastructure. Zhang noted that blockchain and compliant stablecoins can reduce multiple intermediary banks in traditional cross-border payments, enabling near-real-time fund settlement. On-chain stablecoin transactions reached approximately $33 trillion in 2025, yet the proportion actually utilized for the real economy remains low. The key focus for the next phase is addressing compliance, security, and enterprise adoption to let stablecoins truly enter cross-border trade.

At the "AI × New Finance – Innovation Global Tour Hong Kong Stop & Yangtze Star Cluster Plan Launch Conference", Zhou Chenggang, Chief Executive Officer and Executive Director of China Taiping Investment Management (Hong Kong) Co., Ltd., stated that there is an essential difference between RWA and on-chain native assets such as Bitcoin. The token of an RWA is not an asset in itself, but merely an on-chain voucher for off-chain real assets. Therefore, RWA will not break away from the traditional financial system just by going on-chain. Zhou pointed out that issues including how asset rights are established and enforced, who bears responsibility for asset losses, and how information is continuously disclosed still require support from the traditional financial and legal systems. Web3 is more likely to transform some segments of finance rather than completely subvert traditional finance.

At the "AI × New Finance – Innovation Global Tour Hong Kong Session & Yangtze Starry Plan Launch Conference", Zhang Kaifu, founder of Little Grebe AI and former vice president of Alibaba, stated that the core change in AI over the past year is that models have evolved from chat tools to Agents capable of completing complex tasks continuously. The next stage worth focusing on is that AI will start participating in its own research, training, and inference optimization. Zhang Kaifu noted that this "recursive self-improvement" has already emerged in some studies. If in the future AI can continuously optimize models and training efficiency like top researchers, the pace of technological progress may see another exponential jump, and the next 1 to 2 years could become a key variable for AI development. He also judged that as model capabilities continue to improve, AI Agents will shift from passive response to always-on, active task execution, thereby driving new infrastructure demands in cloud computing, identity authentication, cybersecurity and other fields.

At the "AI × New Finance – Global Innovation Tour Hong Kong Session & Cheung Kong Graduate School (CKGSB) Stars Program Launch Conference", Sudan, assistant professor of finance at CKGSB, stated that as information volume continues to explode, AI will evolve from assisting in information processing to further making decisions and executing tasks on users’ behalf. Once AI begins executing transactions directly, payments and asset rights confirmation will become unavoidable links—exactly where Web3 and digital currencies can play a role, making the integration of AI and Web3 highly certain. Sudan also noted that AI is transforming how enterprises compete. Truly valuable AI products in the future cannot merely complete a small step in a workflow; they must form a business closed loop and take responsibility for final outcomes. As model capabilities and workflows become increasingly commoditized, enterprises’ competitive moats will shift more toward exclusive data, real-world scenarios, and deeply integrated business processes.

Beating AI Express (Dongcha) reports that Thibault Sottiaux, OpenAI’s core product lead, hinted Codex could hit a new growth milestone tomorrow. He noted he was reviewing backend data, saying "there might be a new record worth celebrating" and urging users to "Hold on to your Codex." This comment has been widely interpreted by the community as a signal for the next quota reset. Sottiaux (nicknamed Tibo) has previously granted additional quotas to Codex users after the platform hit user growth milestones: in mid-August, when Codex surpassed 15 million active users, he issued a direct quota reset; a week later, after it hit 20 million, he rolled out claimable quota resets for both Codex and ChatGPT Work users. Following the prior pattern of adding roughly 5 million users each time, this milestone is likely targeting 25 million active users.

According to EmberCN monitoring, the Trump Digital Gold (GOLD) scam team sold all its holdings of GOLD tokens at around 2 PM today, totaling roughly 824.54 million tokens — accounting for 82.454% of the total token supply — for a cumulative profit of approximately 9,784.6 SOL (about $1.01 million). GOLD was created on Solana at 7:38 AM today; related addresses had already amassed over 82% of the token supply via distribution and post-launch buys. Around 9 AM, Trump merchandise partner account @realtrumpcoins1 posted a tweet including GOLD’s contract address, pushing the token’s market cap to a peak of $66 million. At roughly 11:48 AM, the tweet was deleted, and the linked addresses immediately began concentrated selling. GOLD’s market cap crashed from $55 million to $1 million in just 30 seconds, after which the scam operators continued offloading until all their tokens were sold. Currently, GOLD’s market cap stands at around $700,000, a 99% plunge from its peak.