902 summarised stories in Companies & Products, each linking back to the original source. Browse all topics →
OpenAI 2 years ago 37
ChatGPT added features that let users interact with tables and charts while uploading files directly from Google Drive and Microsoft OneDrive. The update enables data analysis workflows without manually downloading and re-uploading files from cloud storage services. Users can now process and visualize data more efficiently within the platform.
OpenAI 2 years ago 8
OpenAI will integrate Reddit's content into ChatGPT and other OpenAI products through a new partnership. Reddit's data will be used to train and improve OpenAI's AI models, with financial terms undisclosed. The deal gives OpenAI access to real-time Reddit discussions and community-generated content for model development.
OpenAI 2 years ago 11
Canva has launched an AI-powered Magic Studio feature for its visual design platform used by 175 million monthly users. The tool automates design tasks like background removal, image editing, and layout suggestions to reduce time spent on manual formatting. Users can now generate design elements and compositions without specialized training, lowering the barrier to creating professional-quality graphics.
The daily briefing
TLDRocket reads all relevant 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.
TLDRocket
8 August 2026
Mistral AI's release of Shieldstral 1.0 3B marks a practical inflection point in content moderation: safety classifiers are shedding their fixed rulebooks. The 3B model takes policies as plain-language prompts at inference time—meaning a single deployment can enforce different standards for different customers without retraining—while matching the performance of a 20B baseline on both text and multimodal content. It runs on a single 16GB GPU and ships under Apache 2.0, making it available for local deployment across vLLM and llama.cpp. For teams managing multiple content policies or operating across jurisdictions, this matters because moderation no longer requires separate model forks or constant retraining cycles.
Read the full briefing →