After becoming an overnight sensation, "Lobster" faces a dilemma.

```

An open-source AI project maintained by volunteers is being overwhelmed by its own success.

According to The Information on April 22, OpenClaw—the open-source AI agent software sweeping the tech world—is at a crucial crossroads: Should it maintain its experimental hacker spirit, or transition into an enterprise-level product?

OpenClaw was created by Austrian programmer Peter Steinberger and has rapidly become one of the fastest-growing open-source software projects. Last month, Nvidia CEO Jensen Huang called it the "operating system for personal AI" at the GTC conference, likening its impact to the revolution Windows brought to personal computers. Tech giants like ByteDance and Tencent have successively launched products based on OpenClaw.

However, after going viral, problems have come one after another.

Giants wary of upgrading

Earlier this month, OpenClaw revealed a tricky problem: Nvidia and several other tech companies are still using versions from the first half of March.

The reason is simple—they’re afraid of things breaking. According to three sources familiar with the matter, these companies worry that upgrading to the new version might cause serious technical failures.

This is not unfounded. In the past month, OpenClaw has seen frequent updates, sometimes launching a new version every one or two days, causing custom agents painstakingly built by users to "stop working." Users have also reported incidents of agents deleting data unexpectedly, and cybersecurity firms have discovered malware targeting OpenClaw.

Currently, some maintainers have begun assisting these tech companies in gradually migrating to the latest version, but according to people familiar with the matter, this process is expected to take considerable time.

Internal divisions: Fast iteration or stability first?

These challenges are sparking strategic disagreements within OpenClaw.

According to informed sources, maintainers are split into two camps: One advocates instituting traditional product management models, establishing predictable update cycles, strict quality assurance mechanisms, and formal coordination processes with enterprise users; the other fears that adopting routine processes and rigid structures will stifle innovation and erode the hacker spirit that OpenClaw was built on.

Last week, OpenClaw specifically declared a "feature freeze week" to focus on improving software speed and stability. Yale neuroscience PhD student Gustavo Madeira Santana, who has been an OpenClaw volunteer maintainer since January, said: "We want to improve stability and make sure people aren’t held back by fear of updating."

But some maintainers believe that, given the rapidly evolving field of personal AI agents, OpenClaw must keep up high-frequency updates to continually innovate—versions from just a few days ago might already need to be completely redone.

Noam Schwartz, co-founder and CEO at cybersecurity company Alice, has long been an OpenClaw user. His observation is incisive: "To understand OpenClaw, you need to understand that its starting point was irregular. When OpenClaw was released, it looked more like a product than an open-source project." This led outsiders to have unrealistic expectations of it.

Schwartz admits that each time he upgrades to the latest version, it usually takes him about an hour to fix various issues. "It’s not like an iOS update," he said, "but this is normal for open-source software."

Cost pressures and outside support

Trouble isn’t just coming from within.

Anthropic announced a rule adjustment this month, dramatically increasing the cost for users to access its models through OpenClaw and other third-party tools. Schwartz crunched the numbers: "What was once a $200-a-month subscription to Anthropic could now become $200 a day."

Since many OpenClaw users prefer Anthropic’s Claude model, this change has hit the user experience directly. Madeira Santana said that OpenClaw is working to help users better transition to using OpenAI’s models.

In terms of external support, Tencent recently agreed to donate to the OpenClaw Foundation; OpenAI also proposed a donation (which Steinberger publicly stated he refused), but OpenAI still assisted with the foundation’s paperwork and gave maintainers access to its Codex coding tools; GitHub provided free services to maintainers; Convex paid the hosting costs of ClawHub.

Shirazi proposed a commercialization idea: OpenClaw could launch a paid "Pro version," sold alongside the open-source version, in order to introduce business incentives to improve the product experience—a path other open-source projects have taken before.

Downloads halved, competitors catch up

OpenClaw’s viral momentum is also fading.

According to data from Node Package Manager (OpenClaw's download tool), OpenClaw’s average weekly download numbers have dropped about half since the mid-March peak, returning to early March levels.

Meanwhile, competitors are closing in. According to ClawCharts, which tracks open-source AI agents, Hermes—an agent tool from Nous Research—has already surpassed OpenClaw in number of recent GitHub contributors this month, and continues to chase it in new code submissions, stars, and other metrics.

Darian Shirazi, General Partner at venture capital firm Gradient, summed it up bluntly: "I don’t understand why anyone who isn’t an engineer or tech enthusiast would still use OpenClaw at this point—there are already plenty of solutions, like Town or custom agents in Notion, that work great." He said he has already switched to a competitor because OpenClaw is "cumbersome to use" and poses security concerns.

The founder’s new identity

Steinberger joined OpenAI in February this year, but according to two sources, OpenAI has given him ample time to continue leading OpenClaw’s development. Meanwhile, he is also actively fundraising for the OpenClaw Foundation announced in February, which will oversee the project going forward.

Currently, OpenClaw’s daily operations still rely on dozens of volunteer maintainers, most of whom juggle the work with their own startups and full-time jobs, using AI programming tools to keep up with OpenClaw’s near-constant software updates.

Open-source projects reaching this point is not without precedent. After Linux was released in the early 1990s, it launched a "long-term support release," focused on small updates and bug fixes for long-term stability, so enterprises could depend on it with confidence. Whether OpenClaw can find its own balance remains to be seen.

Risk Warning and DisclaimerMarkets have risks, investment needs caution. This article does not constitute individual investment advice, nor does it consider the unique investment goals, financial situation, or needs of any particular user. Users should consider whether any opinions, viewpoints, or conclusions in this article fit their specific circumstances. Invest accordingly at your own risk. ```