Chinese regulators have told Moonshot AI, StepFun, and ByteDance to reject US investment without government approval, in response to Meta's controversial acquisition of AI agent startup Manus.
Exa's Highlights feature now generates 500-character summaries that match the accuracy of 8,000-character full text, reducing token consumption by 16x. The upgrade is a model-level improvement that automatically benefits all API calls.
Parallel Web Systems, founded by former Twitter CEO Parag Agrawal, has made its search MCP server free and open to all, no sign-up required.
Meta confirmed it will lay off 8,000 employees starting May 20, part of a broader plan to cut up to 20% of its workforce as it ramps up AI investments and efficiency efforts.
Former ByteDance Seed researcher Zhang Chi reveals the company takes ~6 months for a full AI model training cycle, more than double Google's reported 3 months, and says China is 'still far behind' the US with the gap widening.
OpenAI's Clive Chan says DeepSeek V4's hardware recommendations for chip makers are 'surprisingly mediocre and even wrong' — a stark contrast to V3's highly-praised chapter. He questions the report's advice on power, data transfer, and activation functions.
A leading RLHF researcher warns that proposed US anti-distillation legislation would cripple open-source AI startups and academic research, arguing it could set the field back by six to twelve months.
A former White House official says DeepSeek's V4 model is behind and likely using restricted chips. A Silicon Valley CEO argues it's a genuine open-source breakthrough. The battle over China's AI capabilities is heating up again.
In a rare interview, Xiaomi's head of large language models reveals a 1-trillion-parameter model, a no-deadline team culture, and a prediction that AGI is just two years away.
DeepSeek's open-source V4-Pro-Max set a record on Codeforces, beating GPT-5.4, but fell short on knowledge and long-context benchmarks against Gemini and Opus.
