Daily AI ·
OpenAI Gets Green Light for Global GPT-5.6 Launch; Anthropic Unveils Claude Mind-Reading Lens
OpenAI will release GPT-5.6 Sol, Terra, and Luna globally this Thursday after the Trump administration approved broader access. Anthropic launched Claude Cowork on mobile and web with cloud persistence, extended Fable 5 access for subscribers, and revealed the Jacobian Lens that reads Claude's internal 'J-Space' — a hidden reasoning layer that reduces hallucinations. Meta debuted its first Superintelligence Labs image model, Muse Image, sparking privacy backlash. The US cyber agency CISA is using Anthropic's Mythos to audit government code. Google DeepMind targets July 17 for Gemini 3.5 Pro while DeepSeek's API deadline hits July 24.
15 sources2 min read
OpenAI to launch GPT-5.6 Sol, Terra, Luna globally Thursday The Trump administration has approved the broad public release of OpenAI's GPT-5.6 series, ending weeks of restricted U.S.-only preview. The three models — flagship Sol, mid-range Terra, and fast Luna — will be available worldwide starting July 9. 12
Anthropic brings Claude Cowork to mobile and web with cloud persistence Claude Cowork can now run tasks in the cloud, allowing users to start on desktop and continue via mobile or web after closing their laptop. The update also enables scheduled tasks and hands-free agent operation. 345
Meta launches Muse Image model sparking privacy backlash Meta's Muse Image, built by Superintelligence Labs, can generate and edit images using public Instagram content via @mentions, drawing immediate criticism over consent. The company also previewed a Content Seal detection tool for AI-generated images. 678910
Anthropic extends Fable 5 access for Claude subscribers by five days Subscribers on the Max and higher tiers get an additional five days with Anthropic's most powerful model, Fable 5, as the company adjusts availability tiers amid the upcoming GPT-5.6 launch. 11
Anthropic's Jacobian Lens reveals Claude's hidden inner monologue Anthropic discovered that Claude maintains an internal 'J-Space' where it processes concepts silently before outputting them. The Jacobian Lens technique allows researchers to read and steer these activations, leading to new training methods that drastically reduce hallucinations. 1213
US cyber agency CISA uses Anthropic's Mythos to audit government code CISA's Attack Surface Evaluation team is leveraging Mythos to scan government repositories for vulnerabilities, already uncovering a large number of bugs. The move signals deepening government adoption of Anthropic's tools despite past tensions. 14
Google DeepMind targets July 17 for Gemini 3.5 Pro as DeepSeek deadlines loom Google DeepMind aims to release the rebuilt-from-scratch Gemini 3.5 Pro on July 17, the same day DeepSeek graduates V4 to stable. Meanwhile, DeepSeek's legacy API alias deadline hits July 24, forcing developer migration. 15
Daily AI
The most important AI & large language model developments — sourced and cited every morning.