Revolut reports a 90% cost cut by moving AI coding to self-hosted Chinese models, underlining a shift in value from weights to deployment, fine-tuning, and governance. Hugging Face and ByteDance signal that openness is becoming a strategic and economic battleground.
Open source AI is moving into the mainstream. Value is shifting from model weights to deployment, fine-tuning, governance, and TCO. Pressure from Chinese open models pushes the debate toward sovereignty and self-hosted options. An analysis of how inc...
Hugging Face CEO Clément Delangue puts his P(Doom) close to zero and points instead to AI built by a few players behind closed doors. The position reframes danger as power concentration and low inspectability, not the model itself. For teams evaluati...
Open source models promise lower costs and more control, but they raise an uncomfortable question: where does value come from for those who use them? Hugging Face's policy chief frames the risks, trade-offs, and the Chinese challenge.
ByteDance releases Axion 1.0, a 35B agentic LLM with a complete training recipe: two supervised fine-tuning stages followed by reinforcement learning, without mid-training, DPO, or OPD. Beyond weights, it publishes data, pipelines, prompts, and code....
The British fintech shifts part of its development workload to a self-hosted Chinese model and reports 90% savings compared with Claude Code, reopening the debate on TCO, data control and local hardware.
British fintech Revolut has reportedly cut AI-assisted coding costs by a factor of ten, dropping Claude and moving inference to self-hosted infrastructure. A signal for teams evaluating LLM service TCO and control over code flows in regulated environ...
