Oct 11 AI briefing: Grok Bot can now compare prices and place orders for you, and Qiyao, the time-series model built by Huawei with East China Normal University, has topped two international leaderboards — the two stories to read first from the past 24 hours. Also inside: Galaxea's dual-arm platform opens for orders, product updates from Codex and ElevenLabs, and Nicolas Cage's refusal to sign an AI waiver. All six items below were published or took place on October 10-11, 2026 and can be traced to primary sources; stories already covered in standalone articles today, such as Nvidia's talks with Reflection AI and Satya Nadella's emergency-brake proposal, are not repeated.
Grok Bot can shop for you; Musk says a photo of your credit card is enough
On the evening of October 10, Elon Musk posted on X that to buy something you only need to take a photo of your credit card, drop it into the chat, and Grok Bot will search the web for the best price and place the order. On October 11 he posted that he had just had Grok Bot research the most fun Lego Star Wars kits for young kids and order them, calling it just that easy. The demo quickly drew safety pushback: Tibo Sottiaux, who leads Codex and ChatGPT at OpenAI, replied with an AI-generated fake credit card bearing his own name, hinting at the risk of handing card details to a chatbot. One user said Amazon has blocked Grok Bot's access; Musk replied that Grok would shop another way and that it had just ordered from Lego's own site. A caveat matters here: so far there are only Musk's demo posts — xAI has published no formal feature description, and payment authorization, liability for wrong orders and how card data is handled are all unexplained. Agent shopping is moving from helping you choose to paying for you, but until the rules are spelled out, handing payment credentials to a chat box is a demo, not a practice to copy.
Huawei's Qiyao tops two time-series leaderboards
On October 11, according to Huawei Cloud's official announcement, Qiyao — an Ascend-native time-series pretrained model jointly developed by Huawei Cloud's industry model team, the 2012 Application Scenario Innovation Lab and East China Normal University — topped both the TIME-leaderboard and the Pretrained category of GIFT-Eval. Built and optimized for training and inference on the Ascend ecosystem, Qiyao can be used directly for zero-shot time-series forecasting in scenarios such as grid load forecasting, industrial equipment anomaly detection and financial trend analysis, and it supports low-cost fine-tuning. Time-series forecasting is less visible than chat, but it underpins energy dispatch, industrial operations and weather services; topping public benchmarks puts a domestic time-series foundation model in the global first tier, and the next thing to watch is how it holds up on real operational data.
Galaxea's LEMO dual-arm development platform opens for orders
On October 11, Galaxea announced on its official account that the LEMO dual-arm development platform is now available for order. LEMO is an all-in-one platform combining dual-arm hardware, isomorphic teleoperation, multimodal data collection and an open development toolchain, covering the full workflow from data collection and model training to deployment and validation on the robot, aimed at embodied AI research and application development. For robotics teams, data collection and teleoperation are usually the most labor-intensive steps; a platform that bundles hardware, capture and tooling sells exactly the shorter path from unboxing to experiments. Its real delivery pace and developer feedback will only become clear once the first orders ship.
Codex's Composer predictions open to all Pro users
On October 10, OpenAI's Vaibhav Srivastav announced on X that Codex's Composer predictions are now available to all Pro users. The feature reads the whole conversation and project context to predict and suggest the prompt you are most likely to send next — the coding agent hands you a draft before you ask. He also shared his own usage data: over the past two weeks it correctly predicted 36% of his next actions, and another roughly 11% were usable with slight edits. Prompt prediction is a bet on understanding a developer's workflow; if the hit rate holds at this level, a real share of everyday prompt typing disappears.
ElevenLabs launches Brand Kit so generated content follows brand rules automatically
On October 10, ElevenLabs announced on its official account that Brand Kit is live in ElevenCreative. Add your logo, fonts and brand colors once, and everything you create on the platform applies them automatically; in the official demo, asking it to build a launch campaign for a new product produced on-brand output directly. Style drift in AI-generated content has been a standing headache for business users, and brand kits turn brand alignment from after-the-fact rework into a generation-time default. ElevenLabs also announced the opening of its Singapore office the same day, as its creative tooling push continues.
Cage refused Spider-Noir's AI waiver and says he likely won't work with Amazon again
On October 10, at a Spotlight panel at New York Comic Con, Nicolas Cage said he had not signed the AI waiver Amazon put in front of him for Spider-Noir, calling Amazon a very AI-friendly company that has put up to 50 billion dollars into OpenAI, while he is not an AI-friendly actor, and that he probably will not work with the studio again. Variety, Deadline and TheWrap reported the remarks. Spider-Noir premiered in May, earned 11 Emmy nominations and drew 2.6 billion minutes viewed in its first six weeks, yet its second season was canceled last month. Cage opposes letting AI manipulate performances, arguing it replaces integrity and artistic truth with commercial interest; the tug-of-war between actors and platforms over AI authorization has now been aired on a comic-con stage.
