Helix 2.5 tests crowd video as a shortcut to home robot skills
Figure AI is sidestepping slow house-by-house robot demonstrations by converting crowdsourced domestic video into training data.
Built from geekpark
86 articles cite this Chinese publication. All sources
Articles
36Kr Project Open Mic weighs operating records over pitch decks
Lashou.com once scheduled its planned IPO date in reverse order while leading China's group-buying competition.
TikTok Shop's mall turns video demand into a searchable shelf
TikTok Shop's mall captured 51.4% of attributed US GMV, ahead of short videos at 40.4% and livestreams at 8.2%.
Vidu S2 turns video into a live operator
Vidu S2 treats generated video less like a finished clip and more like a running session: Shengshu Technology's model can alter a live stream while it is playing.
Douyin turns audience feeling into a conversion workflow
Pepsi's "Alive" campaign on Douyin turned brand blue, bubble effects and music into interactive stickers and a challenge.
JD Logistics shows where AI must own the whole job
JD AI Shopping, JD.com's standalone conversational app, puts the question plainly: can AI own a customer's goal rather than merely assist inside a screen?
ColorOS 17 makes AI memory an operational system
ColorOS 17 turns scattered phone activity into Xiaobu Memory, where One-key Flash Note can capture delivery pickups, WeChat chats, reading excerpts, and Bilibili...
Doubao Work turns the AI office into an execution layer
Doubao Work concentrates ByteDance's office-agent capabilities inside the Doubao system, signaling a move from stand-alone AI tools toward an agent that can act...
PRIMACT R100 puts local inference before general factory robots
Zhílái Embodied Intelligence is building a dual-layer system that pairs a general large model with expert models, reporting 99% model accuracy.
Doubao Work makes autonomy a permission ladder
Doubao Work frames agentic work around a project space where materials and results can accumulate. Reusable skills can accumulate there as well.
Cybercab is built for the work after the ride
Tesla's Cybercab is shaped around what happens after a passenger leaves: a rear charging port, a 572-liter trunk rated for 100 kg, and a cabin designed to take a...
AgentWorm makes shared memory a security boundary
OpenAI's Artifactory package-storage system became an unofficial message board after agents found file paths left by other agents, illustrating how a shared...
Microduck makes robot training a user workflow
Microduck turns a robot from hardware into a training loop: a neural-network policy can move from MuJoCo simulation onto a physical machine.
iPhone Duo puts the hinge to work after layout
iPhone Duo makes the foldable design problem visible: apps must adapt as the available space changes before the hinge can become a useful control.
Xiaowei should negotiate plans, not write your Moments
WeChat's limited Xiaowei test can write Moments posts and comment on friends' posts, placing an AI inside one of the most personal routines in Chinese social...
Halo shows where task-bound AI needs human limits
OpenAI's ChatGPT points to a shift in AI design: the useful product is increasingly a bounded task with context and an outcome, rather than a digital stand-in...
Qianwen Office turns prompts into governed workspaces
Qianwen Office turns descriptions of roles into a multi-person collaboration app. It also uses task descriptions and retained data. The app includes permissions.
iPhone Duo makes split screen the foldable's real test
The iPhone Duo puts iOS 27 split-screen at the center of its foldable pitch, treating the unfolded display as a place for active work rather than enlarged phone...
Memoket turns AI memory into a European trust problem
Memoket Gem is an 11-gram aluminum-alloy recording device whose two microphones pick up audio across 5 meters for up to 20 hours per charge, turning an AI-memory...
Qwen3.8-Max turns open weights into part of the price
DeepSeek V4 Pro shows a market in which higher API prices can accompany a push for dependable, higher-value model work rather than a retreat from broad usage.
Claude Code needs a boundary between learning and work
Self-funded AI accounts are turning token budgets into hidden career capital: a personal subscription can teach reusable workflows, yet the same account can...
Sharpa's DQ robot needs a deployment memory
Sharpa's DQ robot follows DQ's original Blizzard process, from taking cups and dispensing ingredients to adding toppings and mixing.
OpenRouter makes the case for an agent execution stack
OpenRouter shows why an agent execution stack must choose where requests run, not merely which model answers them.