一起读外刊|OpenClaw热潮背后藏着开源AI的终极困境,大概率走不远-夜雨聆风

一起读外刊|OpenClaw热潮背后藏着开源AI的终极困境,大概率走不远

【写在前面】本文复盘了开源AI项目OpenClaw的爆红现象，指出OpenClaw因打破AI巨头垄断的草根属性走红，却存在安全隐患、体验不佳、进展缓慢等短板。作者认为，用户其实更偏爱成熟易用的闭源产品，叠加大众AI焦虑，OpenClaw难长久出圈，难以颠覆头部商用AI格局。

2026年3月9日发布于《瞭望镜资讯》

这种场景，我们早已在各类科技浪潮中见过无数次……

“相较于谷歌、OpenAI等大型实验室推出的人工智能智能体服务，开源属性让OpenClaw迅速在科技行业走红。但实际上，这款工具仍存在不可控性，潜藏着重大安全风险。不过，业内社群将其视作一场草根革新运动、一项崇高的探索，认为它打破了头部AI企业少数人垄断行业的格局，为行业发展提供了新出路。”

读完这篇报道，我的思绪不禁回溯到科技行业的数个发展阶段。近些年，“Web3.0”的闹剧历历在目，而短暂火爆的NFT热潮更是最典型的缩影。这场热潮起初初衷纯粹、初心本真，但不出所料，很快就被大批投机逐利者裹挟、利用。随着各类骗局频发、项目频频跑路，整个赛道彻底走向衰败。

相比之下，OpenClaw浪潮显然正当得多，也是整场人工智能革命中衍生出的有趣、且在某种程度上顺应趋势的分支。不少人对AI技术满怀期待，却对大型科技企业（以及估值数千亿、自诩“初创企业”的OpenAI、Anthropic、xAI等头部AI巨头）掌控行业、垄断未来的现状心存戒备。OpenClaw恰好填补了这份市场诉求，成为化解行业垄断焦虑的一剂良方，即便历经数次滑稽的品牌更名，依旧站稳了脚跟。

但这场浪潮，让我更多联想到早年社交网络崛起的时代。当年脸谱网取代聚友网成为行业巨头后，开源社交平台OpenSocial应运而生，试图打破其垄断格局。

和当年的OpenSocial一样，我对OpenClaw浪潮的持久力存疑，它大概率只会沦为一段小众新奇的行业谈资，无法真正长久发展、形成规模影响力。这个观点或许会引发很多人的反感，因为这类开源信仰，往往带有近乎宗教般的执念。我认识且欣赏不少投身这场浪潮的从业者。从理论层面来看，大众天然推崇“开源”，认为开源代表着胜利，代表着权力归属于大众，而非企业巨头。

但现实中，开源模式一次次被封闭、成熟的商业化体系碾压。虽非绝对，但已然是常态，在AI这种高风险、大体量的行业赛道中更是如此。社群共建的开源项目，或许开局势头迅猛，最终却往往陷入理念割裂、进展迟缓的混乱僵局。

此外，各大头部AI企业，大概率已针对OpenClaw的热门功能布局了对应的竞品方案。即便OpenClaw赚足了行业热度，普通用户也不会专门购置苹果迷你主机来搭建私有运行实例。更关键的是，绝大多数用户根本不适合自主运行OpenClaw智能程序，极易引发各类安全事故。

再加上用户始终偏爱简洁易用的成熟产品，OpenClaw的结局早已注定，它无法打造出属于自己的AI智能社交生态。

唯一的不确定因素，源于大众对AI行业整体的抵触情绪。纵观科技发展史，除谷歌眼镜之外，几乎没有任何一项技术，能像AI这样引发大众天然的戒备与迟疑。尤其是美国市场，作为全球AI技术的前沿阵地，当地几乎所有AI相关报道，都围绕着“AI取代人类岗位”的焦虑展开。

但我依然认为，大众不会将OpenClaw视作破解AI行业困境的答案。当下，美国政府与Anthropic的冲突风波，反而让ClaudeAI顺势出圈，大众逐渐将Anthropic视作更具“伦理底线”的AI服务商，这份认知，也缓解了不少用户对商用AI的初始偏见与质疑。

随着AI革命持续推进，我始终在思考，人类最终是否会固定选择一款专属AI工具、实现长期复用。未来或许会形成“工作用一款AI、娱乐用一款AI”的格局，但各大头部AI企业正持续完善全场景功能，因此，用户最终固定选用单一AI产品的可能性更大。

不过，倘若深耕OpenClaw赛道的从业者被迫依附于巨头体系，或许会掀起反抗浪潮，转而追求随心所欲的个性化AI体验，恰似某部荒诞派电影《龙虾》的剧情一般。

By M.G. Siegler,Spyglass, Published Mar 9, 2026

It feels like we’ve seen the OpenClaw movie before, many times…

“OpenClaw has quickly become popular in the tech industry for being open-source, in contrast with AI agent services from big labs like Google, OpenAI, and others. Practically, it’s still an unpredictable tool that can pose major security risks. But this community sees it as agrassroots crusade and a noble pursuit, offering an escape hatch from an industry controlled by a handful of people at leading AI companies.” Reading this report, I’m thrown back in time to a few different eras of tech. In recent times there was all the silliness around “web3” – and the brief “NFT” craze was perhaps most symbolic of the situation. It undoubtedly started out as earnest and innocent enough, but quickly – as expected – it was overtaken by a lot of grifters trying to make quick bucks. And the whole thing sank under the weight of the various schemes as rugs were pulled.

This OpenClaw movement certainly seems far morelegitimate, and it feels like an interesting and in some ways natural offshoot of the broader AI revolution. Some people want to be excited about AI, but are wary of Big Tech (and “Big AI” – the OpenAIs, Anthropics, and xAIs valued in the hundreds of billions of dollars as “startups”) controlling the situation and potentially the entire future. OpenClaw has slipped right in as an ointment for that ailment – to the point where it has even survived those comical branding shifts. Still, I’m reminded more of an earlier era, when social networks were rising in prominence and power. Once Facebook overtook MySpace, in came OpenSocial to try to combat their inevitability.

As was the case with OpenSocial, I’m skeptical that this movement lasts and gets any real legs beyond these sort of fun, quirky stories. This sentiment will make people angry, because such things tend to almost be more religious in nature. And I know and like a lot of the people involved in this particular movement. On paper, people love “open”. It naturally sounds “winning” to the masses. Power to the people and not the corporations, and all that. But in reality, we see open getpummeled by closed, perfected systems time and time again. Not always, but often. And especially, when the stakes and the scale are at AI levels. The “open” stuff built by a community may start strong but often degrades into a mess of contradictory ideas and simply slow progress.

And I also imagine that each of those “Big” players has their own answers for some of the elements of OpenClaw that prove popular. While the OpenClaw movement is fun, headlines aside, the masses are simply not going to be rushing out to buy Mac minis on which to run their owninstances. And again, most people should not be running their own OpenClaw bots, as it’s a recipe for various security disasters. That mixed with the aforementioned tendency for people to gravitate towards good, easy-to-use products makes the end game fairly obvious here. And it’s not a social network for OpenClaw bots.

The onlywild card may be the aforementioned feelings about AI in general. Again, more so than perhaps any other technology in the past – aside from maybe Google Glass – there’s seemingly a real natural hesitancy here. Certainly in the US market, which is obviously at the forefront of the technology and at the same time, nearly every story you read about AI is about the fear over job displacement. But I still have a hard time thinking that cohort of people is going to gravitate towards OpenClaw as the answer. In a way, we may be seeing a shift towards Claude itself because of the blow up between the government and Anthropic. As a result of that, people may be starting to see Anthropic as the more “ethical” solution in AI, which may help some over that initial hump of skepticism.

As we go further down the path of this AI revolution, I continue to wonder if and when we’re all going to settle down with our one true AI. It’s possible we end up in a world where we use one for work and one for play, as it were, but given that the aforementioned big players are increasingly trying to do it all, I think picking one seems more likely. Though if the ClawCon attendees are forced into such relationships, perhaps they’ll revolt and instead demand to be turned into an animal of their choosing. Just like the plot for a certainabsurdist movie.

原文链接

https://spyglass.org/openclaw-versus-closed-claude/

重点词汇

grassroots/ˌɡrɑːsˈruːts/ adj. 草根的；民众的

○ involving the ordinary people in a society or an organization

○ e.g. grassroots support 基层民众的支持

crusade/kruːˈseɪd/ n.（长期坚定不移的）斗争，运动

○ a long and determined effort to achieve something that you believe to be right or to stop something that you believe to be wrong

○ e.g. to lead a crusade against crime 领导打击犯罪活动的运动

escape hatch安全舱口；逃生舱口；紧急出口

○ a door or hatch that can be used to get out of something in an emergency

○ e.g. the escape hatch of a submarine 潜艇的紧急出口

earnest/ˈɜːnɪst/ adj. 非常认真的；真诚的

○ very serious and sincere

○ e.g. an earnest young man 非常认真的年轻人

grifter (grifters)/ˈɡrɪftə(r)/ n.（骗取钱财等的）骗子

○ a person who tricks people into giving them money, etc.

○ e.g. She fell for a charming, fast-talking grifter. 她相信了一个魅力十足、花言巧语的骗财者。

scheme(schemes)/skiːm/ n. 阴谋；诡计；计谋

○ a plan for getting money or some other advantage for yourself, especially one that involves cheating other people

○ e.g. They hatched a scheme to rob a Monte Carlo casino. 他们策划了一个抢劫蒙特卡洛赌场的计划。

legitimate/lɪˈdʒɪtɪmət/ adj. 合法的，正当的，法律允许的

○ allowed by law

○ e.g. The army must give power back to the legitimate government.军队必须将权力交还给合法政府。

offshoot/ˈɒfʃuːt/ n. 分支；（尤指）分支机构

○ a thing that develops from something, especially a small organization that develops from a larger one

○ e.g. commercial offshoots of universities 大学附属的商业机构

wary/ˈweəri/ adj.（对待人或事物时）小心的，谨慎的，留神的，小心翼翼的

○ careful when dealing with somebody / something because you think that there may be a danger or problem

○ e.g. Be wary of strangers who offer you a ride. 提防那些主动让你搭车的陌生人。

ointment/ˈɔɪntmənt/ n. 药膏；软膏；油膏

○ a smooth substance that you rub on your skin to make a wound or painful place get better or stop hurting

○ e.g. antiseptic ointment 抗菌软膏

ailment/ˈeɪlmənt/ n. 轻病；小恙

○ an illness that is not very serious

○ e.g. childhood / common / minor ailments 儿童期／常见／轻微小病

branding/ˈbrændɪŋ/ n. 品牌创建

○ the activity of giving a particular name and image to goods and services so that people will be attracted to them and want to buy them

○ e.g. He sees contemporary branding as ‘service with personality’. 他将当代品牌视为“个性化服务”。

inevitability/ɪnˌevɪtəˈbɪləti/ n. 不可避免；不能防止

○ the fact that something cannot be avoided or prevented

○ e.g. the inevitability of death 死亡的必然性

pummel(pummeled) /ˈpʌml/ v. 连续猛击；反复拳打；捶打

○ to keep hitting somebody / something hard, especially with your fists (= tightly closed hands)

○ e.g. He pummelled the pillow with his fists. 他用双拳不停地捶打枕头。

degrade (degrades)/dɪˈɡreɪd/ v. 降低…质量；使恶化

○ to make the quality of (something) worse

○ e.g. Scratches on a camera lens will degrade the image. 相机镜头上的划痕会影响成像效果。

instance (instances)/ˈɪnstəns/ n. 实例

○ In computer science, an instance or token (from metalogic and metamathematics) is a specific occurrence of a software element that is based on a type definition. When created, an occurrence is said to have been instantiated, and both the creation process and the result of creation are called instantiation.

be a recipe for something (it’s a recipe for)可能会造成某事（常指坏的结果）

○ to be likely to cause a particular result, often a bad one

○ e.g. She said that five small boys on skis was a recipe for disaster, not a holiday. 她说五个来滑雪的小男孩就不是来度假的，而是来闯祸的。

aforementioned/əˈfɔːmenʃənd/ adj. 前面提到的；上述的

○ mentioned before, in an earlier sentence

○ e.g. The aforementioned person was seen acting suspiciously. 有人看见前面提到的那个人行动可疑。

gravitate towards被吸引到；受吸引而转向

○ to move towards somebody / something that you are attracted to

○ e.g. Many young people gravitate towards the cities in search of work. 许多年轻人被吸引到城里找工作。

wild card猜不透的人；不可预料的事

○ a person or thing that could affect a situation in a way that cannot be predicted; an unknown or unpredictable factor

○ e.g. Taxes are the wild card in this election. 在这次选举中税收政策是无法预知的因素。