Apple exposes Siri AI to external developers while Anthropic tackles verifiable model routing.
Today's updates emphasize developer access and platform transparency: Apple has opened Siri AI to developers ahead of public rollout, Anthropic has open-sourced an experimental mechanism to cryptographically audit model inference, and Meituan has introduced a trillion-parameter multimodal agentic model.
Apple Opens Siri AI to All Developers
SourceBuilders can now integrate core app workflows directly into Siri AI, though server-side daily limits mean critical features are building on a surface that will eventually carry unannounced usage fees.
Anthropic Releases FlashProve for Provable Model Inference
SourceOffers the first open-source cryptographic proof verifying that an inference response came from the claimed model weights, protecting developers from unannounced silent downgrades by API providers.
Meituan Unveils LongCat-2.5 Native Multimodal Agent Model
SourceDelivers a 1.6T-parameter, 1M-token context model activating ~48B parameters per token, specifically optimized for builders running long-horizon autonomous coding and GUI-driven agents.
OpenAI Integrates Work Features into ChatGPT and Doubles Pro Limits
SourceDoubled usage allowances and embedded enterprise collaboration tools lower the barrier and cost for small teams utilizing ChatGPT as a primary workspace agent.
Get tomorrow's digest the moment it ships.