Every AI story that matters — and the intelligence behind it.
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.
GPT-5.2 proposed a new formula for gluon amplitudes in theoretical physics that researchers subsequently proved and verified. The formula describes interactions in quantum chromodynamics, a field where such formal advances are rare enough to warrant publication in a preprint. The result demonstrates that large language models can contribute to mathematical proof discovery, though the practical applications remain unclear.
OpenAI introduced Lockdown Mode and Elevated Risk labels in ChatGPT to help organizations protect against prompt injection attacks and AI-driven data theft. Lockdown Mode restricts ChatGPT's functionality to prevent attackers from manipulating the system into unauthorized actions or data extraction. Organizations can now identify and mitigate security risks by labeling threats as elevated, enabling better defense strategies against emerging AI vulnerabilities.
OpenAI built a real-time access system that combines rate limits, usage tracking, and credits to manage continuous access to Sora and Codex. The system handles variable demand while maintaining service stability across both products. This approach allows OpenAI to scale access without simply removing restrictions, instead distributing availability based on tracked consumption and credit balances.
OpenAI released GABRIEL, an open-source toolkit that converts qualitative text and images into quantitative data using GPT models. The tool is designed to process large volumes of research materials, enabling social scientists to analyze data at scale rather than manually coding smaller samples. This allows researchers to expand the scope of qualitative studies by automating the conversion of unstructured material into structured datasets.
Codex and Claude generated production-ready CUDA kernels for a video diffusion pipeline and a language model using a new agent skill that packages GPU optimization knowledge. The RMSNorm kernel achieved 1.88x speedup on isolated benchmarks for LTX-Video and 1.94x for Qwen3-8B, translating to 6% and measurable end-to-end improvements when combined with torch.compile. Users can now install the skill, prompt an agent to build optimized kernels, benchmark them against PyTorch baselines, and publish pre-compiled binaries to the HuggingFace Kernel Hub for one-line loading by others.
Every AI story that matters,
in your inbox by 8am.
TLDRocket reads 60+ sources, removes duplicate coverage, and summarises the
day in two minutes. Follow companies and topics for alerts, or get the
briefing in Slack. Free, no spam, unsubscribe anytime.
