Fu Sheng's AI Agent 'Lobster 30K' Hit by Reskin Allegations

Fu Sheng, the prominent CEO of Cheetah Mobile, is facing a wave of criticism after his newly promoted AI Agent product — dubbed 'Lobster 30K' (龙虾三万) — was accused of being little more than a repackaged open-source project. The controversy erupted shortly after Fu Sheng took to X (formerly Twitter) to tout the product's migration to DeepSeek V4 Pro, highlighting dramatic cost savings compared to Anthropic's Claude Sonnet.

The incident has reignited a broader industry debate about the thin line between building on open-source foundations and simply 'wrapping' existing tools with a new interface.

Key Takeaways

• Fu Sheng promoted Lobster 30K's switch to DeepSeek V4 Pro on X, citing a 17x cost reduction versus Claude Sonnet
• Critics allege the product is a 'reskinned' version of an existing open-source AI Agent framework
• DeepSeek V4 Pro pricing sits at roughly $0.87 per million tokens, compared to Sonnet's $15 per million tokens
• The controversy highlights growing tension in China's AI ecosystem around originality vs. open-source repackaging
• 'Wrapper' accusations have become increasingly common across both Chinese and Western AI startups
• The debate raises fundamental questions about what constitutes genuine product innovation in the AI Agent space

Fu Sheng's Bold Cost Claims Spark Scrutiny

Fu Sheng's original post on X was enthusiastic. He declared that Lobster 30K had 'fully switched to DeepSeek V4 Pro,' claiming the experience was indistinguishable from — and in some tasks superior to — Anthropic's Claude Sonnet. The cost differential was the centerpiece of his pitch: at roughly $0.87 per million tokens, DeepSeek V4 Pro undercuts Sonnet's $15 per million tokens by a factor of 17.

'No more token anxiety,' Fu Sheng wrote, framing the migration as a breakthrough for cost-conscious AI Agent deployment. The post was clearly designed to position Lobster 30K as a practical, affordable AI solution.

But the promotional effort backfired. Within hours, developers and industry observers began raising pointed questions about the product's underlying architecture, alleging that Lobster 30K bore striking similarities to well-known open-source AI Agent frameworks.

'Reskin' Accusations Gain Traction Online

The core allegation is straightforward: critics claim that Lobster 30K is essentially a 'wrapper' — a product that takes an existing open-source codebase, adds a branded user interface, and markets it as a proprietary solution. In Chinese tech circles, this practice is commonly referred to as '套壳' (tào ké), which translates roughly to 'putting a shell on' an existing product.

Several developers on Chinese social media platforms pointed to specific code patterns and architectural choices within Lobster 30K that they say mirror popular open-source AI Agent projects. While the exact open-source project at the center of the allegations has been debated, the sentiment is clear: many believe Fu Sheng's team did not develop sufficient original technology to justify marketing Lobster 30K as a proprietary product.