AI News AI资讯 3mo ago Updated 2mo ago 更新于 2个月前 85

Neolithic NewClaw: Integrated AI Solution, Zero Barrier to Becoming an Autonomous Vehicle Commander | 2026 AI Partner · Beijing Yizhuang AI+ Industry Conference 新石器NewClaw:AI一体化解决方案,零门槛当无人车指挥官| 2026AI Partner·北京亦庄AI+产业大会

Neolix, a leading autonomous delivery company, has launched its proprietary AI Agent, NeoClaw, to tackle the next frontier in scaling unmanned logisti 新石器公司推出全栈自研的AI Agent“NeoClaw”,旨在解决无人车规模化运营中的**车队管理瓶颈**。该产品通过自然语言交互,将**单人管理效率提升10倍**以上,实现“一人管千车”。此举标志着无人配送行业竞争焦点从自动驾驶技术本身,转向以**AI驱动的高效运营与商业模式**(如RaaS)创

85
Hot 热度
90
Quality 质量
80
Impact 影响力

Analysis 深度分析

One person. One phone. One sentence. That’s the command-center fantasy NeoClaw is selling for autonomous fleet management. Forget the sleek dashboard or the army of remote operators; the future of logistics, according to Neolix, is a group chat with a fleet of robots. Their new AI Agent, NeoClaw, promises to turn fleet management from a specialized job requiring a team into something as intuitive as sending a voice message. It’s a bold claim, and one that feels less like an incremental update and more like a philosophical pivot for the entire autonomous vehicle industry. The technical hurdles of self-driving, it suggests, are becoming yesterday’s problem. The new frontier isn’t making the cars drive—it’s making them be useful at scale.

Neolix isn’t pulling this out of thin air. They’ve spent seven years climbing a ladder with three clear rungs: building a factory capable of mass-producing Level 4 vehicles, navigating the regulatory maze to get China’s first commercial license, and finally, this year, scaling to a fleet of 10,000 vehicles. That’s the context behind NeoClaw. It’s not a standalone gimmick; it’s the necessary operating system for a company that has, for better or worse, made the bet that autonomous delivery isn’t a tech demo but a logistics utility. Their “triple jump” is real, and it grants them a credibility that purely software-focused AI startups lack. You can’t talk about managing a fleet you’ve never actually built and deployed.

The pitch is seductive: a tenfold increase in management efficiency, from one person overseeing ten vehicles to one overseeing a hundred. The “chat to command” interface is the killer feature. Instead of parsing complex telemetry on a screen, a manager could theoretically say, “Prioritize restocking the downtown medical center, reroute vehicles 42 and 57 to handle the spike, and run a diagnostic on the rest during off-peak hours.” This frames AI not as an autonomous decision-maker but as an ultimate force multiplier for human judgment. It’s a deliberate move away from the “black box” narrative of full autonomy towards a more palatable “co-pilot” model. This is smart. In a world terrified of ceding complete control to algorithms, positioning the human as the conductor, even if the orchestra plays itself, is a potent psychological and regulatory play.

But let’s inject some cold water into this warm, efficient future. “Zero barrier to entry” is marketing hyperbole. Managing a thousand physical robots roaming public streets, even via natural language, is not like managing a swarm of Amazon warehouse bots. It involves navigating local traffic regulations, physical maintenance logistics, battery swaps, cargo loading protocols, and the inevitable edge case where a vehicle gets stuck in a festival crowd or a snowstorm. The AI agent is handling routing and fleet optimization, but the messy, tangible world of logistics doesn’t disappear. The leap from “professional operation” to “just talk” glosses over the substantial, and frankly unglamorous, human expertise still required to keep such a system resilient. One person might issue the commands, but you’ll still need a dispersed, human maintenance crew to execute on the physical consequences of those commands.

Furthermore, Neolix’s technical philosophy is pragmatically, almost ruthlessly, focused on cost. Their “No-Map” approach, end-to-end learning from abundant real-world data, and deep logistics know-how are all engines of TCO reduction. This isn’t the pursuit of perfect, city-wide autonomy like Waymo’s initial ambition; it’s the pursuit of a specific, profitable use case—commercial delivery within defined areas. Their acknowledgment that the real methods are just “simulation, end-to-end, and reinforcement learning,” dismissing fancier terminology as “marketing,” is refreshingly candid. It also reveals a core tension: they are building a commodity. The value isn’t in a proprietary secret sauce of driving logic, but in the integrated stack—the hardware, the software, the data from 150 million kilometers, and now, the management layer—that allows them to deploy cheaply and at scale.

This leads to their RaaS (RoboVan-as-a-Service) model, which is perhaps the more radical idea than the AI agent itself. Selling a vehicle is a transaction. Renting mobility-as-a-service is a recurring relationship, one that aligns the company’s success directly with the vehicle’s utilization rate. A 24/7 machine that only incurs marginal cost the more it runs is the holy grail of any capital-intensive business. NeoClaw is the tool that makes RaaS viable for customers. It lowers the operational skill floor, allowing a small logistics firm to “rent” 50 bots and manage them as easily as it manages a couple of human couriers. It’s a classic platform play: own the tools that lower the barrier for others to use your infrastructure.

The global expansion to 20 countries, with a focus on the Middle East, hints at where they think this model is most immediately fertile: places with rapid urbanization, a appetite for tech, and perhaps less legacy infrastructure or regulatory inertia. Building a network of 10,000 vehicles in one region is a statement of intent to create a new logistics backbone, not just sell individual units.

So, is NeoClaw revolutionary? It’s a critical milestone, but perhaps more evolutionary. It represents the point where the autonomous vehicle industry starts solving its own second-order problems. The first problem was “can it drive?” The second is “can we operate a thousand of them without going bankrupt?” NeoClaw is an answer to the second. It’s the dumb-terminal-to-smartphone moment for fleet management. The real test won’t be in a demo, but in the chaotic streets of a city like Riyadh or Singapore, when a manager’s natural language command interacts with a thousand real-world variables. If it works, it won’t be because the AI is magical, but because it’s the right tool for the brutally practical business of moving stuff around the planet, just a little bit more efficiently. And that, in the end, might be a more significant achievement than the self-driving itself.

管理一千台无人车只需一个人、一部手机、一句话——这个口号在2026年的北京亦庄AI产业大会上抛出时,空气中弥漫着一股熟悉的科技发布会味道。新石器的联合创始人颉晶华站在台上,像推销新手机一样介绍他们的AI Agent NeoClaw,声称能把单人管理效率从十台拉升到一百台以上。但等等,这听起来是不是太像科幻片了?我们真的已经到了“说话就行”的时代,还是又一场精心包装的营销狂欢?

新石器这家公司在无人配送赛道上混了八年,从2019年搞出所谓的“全球第一个L4级别无人车万台量产工厂”,到2021年在北京亦庄拿到第一张无人配送牌照,再到2025年宣称运营万台车队——这三级跳听起来挺唬人。但仔细想想,L4级别自动驾驶在实验室里吹了这么多年,真正落地量产的有多少?万台车队?数据不会说谎,但数据可以包装。1.5亿公里的累计运营里程,数字庞大,可除以万台车,每辆车平均才跑1.5万公里,这规模在物流行业里连个水花都溅不起来。更别提那1500个专利,一半以上是发明专利,但专利不等于产品,更不等于利润。新石器自称是全球无人配送领军企业,可领军到什么程度?在阿联酋、泰国、新加坡部署几个POC(概念验证),就敢喊出今年在中东部署一万台车的目标——这种扩张速度,要么是真有底气,要么是画饼充饥。

技术上,新石器标榜无图技术、端到端大模型、强化学习,这些词儿现在是自动驾驶行业的标配话术。颉晶华自己都说了,行业大佬们讲自动驾驶没那么神秘,方法无非三种:虚拟仿真、端到端、强化学习。没错,抓眼球的词语更多是为了营销。商业自动驾驶的核心是降本,这点他没说错。摆脱高精地图依赖,确实能降低部署成本,因为高精地图的采、建、制、验每个环节都烧钱。但无图技术就真的成熟了吗?在复杂城配场景里,全天候全时段运营,靠的是数据堆积和算法迭代,新石器说“生于物流、长于物流”,有行业Know-how,可物流行业的坑多的是,从装卸到搬运,他们自己也承认现在只解决了运输环节。推出NeoClaw这个AI Agent,宣称像聊天一样指挥车队,但真正的车队管理涉及调度、故障处理、路径优化等一堆琐事,一个Agent能搞定?听起来像把复杂系统简化成玩具——当然,如果目标用户是小白,那“零门槛”可能真能吸引眼球,但专业运营商会买账吗?

商业模式上,新石器搞RaaS(RoboVan-as-a-Service),从去年开始推行,从卖车转向提供即时服务。这思路不错,无人车24小时运营,边际成本递减,理论上能赚到钱。但现实是,无人配送车的维护、保险、充电/换电基础设施,哪一样不是烧钱大户?RaaS模式依赖于稳定的订单流和高效的运维网络,新石器在海外20个国家布局,可每个市场的法规、基础设施、用户习惯都不同,盲目扩张很可能摊子铺太大,收不回来。目标在中东部署一万台车?中东市场对无人配送的需求真有那么大,还是新石器在蹭地缘热点?

回到NeoClaw本身,这AI Agent号称让管理从“专业操作”变成“说话就行”。但指挥车队不是发朋友圈,一句话可能涉及成千上万的变量。如果真这么简单,物流公司早该裁掉所有调度员了。或许NeoClaw在原型阶段能处理简单指令,但面对突发情况,比如车辆故障、交通堵塞或客户投诉,AI能像人一样灵活应变吗?颉晶华强调AI的价值是解放双手,但解放双手不等于消灭专业性——这更像是一种过度乐观的简化,把行业痛点包装成科技噱头。

总的来说,新石器这八年确实做出了些成绩,从量产到牌照到运营,算是国内无人配送领域的先行者。但行业现状是,技术还没完全成熟,商业模式仍在试水,而新石器却急着用AI Agent来画一个更大的蓝图。自动驾驶的“平权”还早得很,瓶颈从技术转向运营,这话没错,可运营的复杂性远超想象。NeoClaw或许能提升效率,但把它吹成“一体化解决方案”,让一人管千台车,这听起来更像是给投资者看的故事。在科技圈,我们见过太多“革命性产品”最后沦为PPT上的传说——新石器这次,是真金不怕火炼,还是又一场华丽的泡沫?时间会给出答案,但眼下,保持怀疑总比盲目喝彩更明智。

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。