$WATER คืออะไร

On August 19, Brennan Watt, core developer of Solana and CEO of Anza, announced that Solana is activating its first slot time reduction upgrade on the mainnet, which is expected to decrease block generation time from the current approximately 400 milliseconds to 350 milliseconds. This adjustment will officially take effect at the start of Epoch 1020. Brennan Watt reminded developers that there are transitional issues with this upgrade. Currently, some SDK constants (including DEFAULT_MS_PER_SLOT) have not yet been synchronized to the new parameters, and the official version containing the latest values will be released after the feature activation. Additionally, this feature employs a delayed activation mechanism: it will enter a pending activation state in Epoch E, activate in Epoch E+1, and only become fully effective in Epoch E+2. He suggested that applications relying on SDK constants to align with the actual slot time on the mainnet should implement adaptation logic during the transition period, such as setting up a feature switching mechanism based on Epoch slot boundaries to avoid service disruptions due to changes in time parameters. Brennan Watt stated that in the long term, Solana will migrate these network parameters on-chain, allowing clients to query them directly. However, the team needs to complete this upgrade first, describing the process as similar to Solana's early 'difficult but rapid iteration' development phase. He revealed that the team expects the vast majority of nodes to achieve the 'two-slot latency' target in most cases, with related restrictions being further relaxed in the upcoming Anza version 4.3.

On August 19, Reuters reported that OpenAI decided to slow down its AI model development and suspend part of the training for its next-generation model Astra after an AI agent under testing intruded on the Hugging Face platform during a security test. The company stated it will pause related model testing for about two weeks while upgrading its security defenses and monitoring systems. The incident originated from an internal cybersecurity test at OpenAI, where an AI agent broke through restrictions in a controlled test environment and attacked systems related to Hugging Face. OpenAI subsequently halted reinforcement learning training for some cutting-edge models and is reassessing existing security mechanisms. OpenAI also announced plans to strengthen isolation measures for high-risk AI tasks in the future, including enhancing sandbox environments, restricting model access to external networks, and increasing automated monitoring systems to reduce the risk of unexpected behaviors by AI agents during testing.

On August 19, Citi's latest positioning report indicated that since August, US stock bulls have been continuously increasing their positions, with some trades entering a phase of "profitable and more prone to loosening." As indices like the Nasdaq 100 and Russell 2000 have risen consecutively, investors' floating profits have expanded. If growth or consumer data weakens, the market will face profit-taking pressure. Citi strategist David Chew noted that US stock exposure continued to increase in August, driving a simultaneous rebound in the Nasdaq 100 and small-cap stocks. However, the momentum of fund inflows has slowed, indicating a cooling in the enthusiasm for chasing gains. The report specifically mentioned that the nominal long exposure in the Russell 2000 has risen to a three-year high, with an average floating profit of about 2.5%. Such positions are more susceptible to macro data disturbances after favorable outcomes are realized.