Every AI story that matters — in your inbox by 8am.
TLDRocket reads 60+ sources, removes duplicate coverage, and publishes a short neutral
summary of every story, linking back to the original. Free, no spam, unsubscribe anytime.
DeepSeek released its first proof submissions for the First Proof math competition, a challenge designed to evaluate advanced AI reasoning on difficult mathematical problems. The submissions represent a research-grade test of the model's capabilities on expert-level mathematics without specifying performance metrics or outcomes. This represents an early data point in assessing how current AI systems handle formal mathematical reasoning compared to human mathematicians.
Hugging Face and Unsloth are offering free credits to fine-tune language models on their Jobs platform, with training speeds approximately 2x faster and 60% less memory usage than standard methods. A 1.2 billion parameter model can be trained on a t4-small GPU for roughly $0.40 per hour, while coding agents like Claude Code can automate script generation and job submission. Users can now fine-tune and deploy small models directly through natural language prompts rather than writing training code manually.
GGML creator Georgi Gerganov and his team are joining Hugging Face to support the long-term development of llama.cpp, the widely-used local AI inference tool. The llama.cpp project will remain 100% open-source with Gerganov maintaining full technical autonomy and leadership, while Hugging Face provides sustainable resources and infrastructure. The collaboration aims to simplify deployment of open-source models locally by creating seamless integration between llama.cpp and Hugging Face's transformers library.
