Arc has released Arc AI Screen Assistant, a floating overlay that reads any screen and offers one-tap AI actions like summaries, read-aloud and Smart Extract. Arc says the tool is available on Android and Mac and is free to start.
10xJoy, created by Philip and Cristian, debuts an early beta to convert business goals into editable briefs and connect clients with builders by email. The service is free for clients and builders; project scope, contracts and payments are negotiated directly with builders.
Ringg reports that its agent platform, built on GPT‑5.6 and other OpenAI models, handles more than 7 million connected calls each month and resolves up to 65% of routine inquiries. The company says migrating some workloads from GPT‑4.1 to GPT‑5.6 cut model costs by about 90%.
Airbnb has broadened engineering and product teams’ access to OpenAI frontier models, including GPT‑6 Astra, via OpenAI APIs and Amazon Bedrock. The company said early use has accelerated engineering work and coincided with a reported 80% rise in shipped features year‑over‑year.
Synthesia created an interactive digital avatar of reporter Dominic-Madori Davis that answers questions only about her article on venture-backed startup fraud. The avatar was made from photos and a two-minute voice sample and uses Synthesia’s video and voice models.
Jango is a testing tool that runs groups of AI agents to simulate multiple users interacting in real time. The product description says each agent has its own browser, account, goals and memory, and that tests produce reports with actions, errors and screenshots.
Floot says Floot MCP connects to Claude and ChatGPT4.3 to produce full‑stack web and mobile apps from a chat. The company says projects include a database, user logins, hosting and a live URL, and that AI usage is billed to the user’s Claude or ChatGPT subscription.
The Transformers team added support for running GGUF (llama.cpp) quantized checkpoints inside transformers, with an initial focus on Apple Silicon and Qwen3.5. Benchmarks reported by the team show performance close to llama.cpp on a MacBook Pro M2 Max.
OpenAI says GPT‑6 now uses an improved prompt caching system that raises cache hit rates and offers discounts of up to 90% on cached input tokens. New tools include the Prompt Caching Dashboard, a diagnostics tool, explicit cache controls and prewarming.
OpenAI introduced GPT‑6 Sol and GPT‑6 Luna as lower-cost variants trained with methods from GPT‑6 Astra. The company reports up to 50% lower API prices versus GPT‑5.6 and presents benchmark and caching improvements it says make the models more cost‑efficient.