谷歌推出了Veo 3.1 Lite 视频生成模型，以 0.05 美元/秒 的低成本提供 720p 分辨率，并同步下调了 Veo 3.1 Fast 的定价。Veo 3.1 Lite 适合成本敏感的高频调用场景，而 Fast 版本则面向需要高质量输出的应用。此举将降低 AI 视频生成门槛，促进市场普及。

Investigator ZachXBT exposed "Rashid bin Saeed" as a fake account that rapidly accumulated followers by posting war-related content, with the goal of promoting the low-cap token CHIBI for market manipulation. Through characteristics such as frequent username changes, abnormally fast verification status updates, and disproportionate follower counts, this account displayed typical fraud indicators. In an extremely fearful market, such manipulation tactics are more likely to succeed.

OpenAI on March 18th launched GPT-5.4 Mini and GPT-5.4 Nano, designed specifically for high-capacity AI workloads. The Mini model is twice as fast as its predecessor, while Nano reduces costs and optimizes real-time conversations. Mini has approached flagship performance on benchmarks and exceeded human baselines, while Nano is positioned as a developer tool with more competitive pricing. Mini is open to users, while Nano is available via API access only, suitable for large-scale applications.

Ethereum founder Vitalik Buterin proposed a new fast confirmation rule that guarantees Ethereum transactions become irreversible within 12 seconds, aiming to improve the existing mechanism that requires 6.4 minutes for final confirmation. This proposal relies on the majority of validator nodes being honest and network latency being below 3 seconds. If these prerequisites hold, it can effectively enhance user experience.

xAI 创办人 Elon Musk 以经典 meme 说明 AI 投资逻辑：「先买一大堆 GPU、中间保持问号、最后获利」。 （前情提要：马斯克怒了！xAI 控告前工程师窃取 Grok 机密，套现 700 万美元跳槽 OpenAI ） （背景补充：马斯克xAI推出「Grok Code Fast 1」编程模型：Cursor、Copilot、Windsurf..限时免费用，速度 规模 价格三突破 ） 那个网友熟悉的，爱讲干话的马投顾又来了！今天身为特斯拉与 xAI 创办人的Elon Musk 在自己的 X 帐号贴文，以他一惯喜好的 meme 模式描绘科技产业的获利想像： 步骤 1：买一大堆 GPU