On April 4, 2026, Boris Cherny, head of Claude Code—an AI programming tool under Anthropic, a prominent U.S. large-scale model company—announced that starting at 12:00 PM Western U.S. time on the same day, the Claude subscription service would no longer include usage quotas for third-party tools like Open Claw.

For most users who rely on both Open Claw and Claude, this adjustment effectively amounts to a prohibition.

The actions of major corporations are often propelled by competition, and Open Claw's meteoric rise has sent shockwaves throughout the tech industry.

This open-source AI agent framework, endearingly dubbed 'Lobster' by developers, has witnessed over 200,000 daily downloads on GitHub. By March, it had amassed over 1 million global deployment instances and garnered more than 250,000 stars, eclipsing all previous records for open-source software projects. It has propelled large language models from merely 'being able to converse' to 'being able to accomplish tasks.' It can directly navigate browsers, read files, invoke APIs, execute scripts, and autonomously dissect and carry out tasks.

More crucially, Open Claw is reshaping the entry point for human-computer interaction. In the traditional internet era, search engines served as the gateway to information; in the mobile internet era, super apps became the portal to services; in the AI Agent era, personal agents are poised to become the unified interface connecting users to all digital services. Data from the National Data Bureau reveals that China's daily Token calls have skyrocketed from 100 billion in early 2024 to 140 trillion in March 2026, marking a thousandfold increase in just two years.

2026 is poised to be the year AI truly takes flight.

This implies that whoever can retain users' 'lobsters' within their own 'shrimp ponds' will seize the initiative in the future ecosystem.

I. 'Lobster' Disrupts Big Tech's AI Strategies

1. Baidu's Comprehensive Strategy

Faced with this historic opportunity, over a dozen major companies, including Tencent, Alibaba, ByteDance, and Baidu, have made intensive strategic布局 (strategic deployments) within a month, officially igniting the 'Hundred Shrimp Battle.' However, each company has chosen vastly different strategic paths. Amidst this chaos, Baidu has once again opted for a 'full-stack technology' approach.

In essence, rather than betting on a single entry point, it's more prudent to pursue a comprehensive layout.

Baidu's core strategy can be encapsulated as the 'Lobster Family Bundle,' with parallel advancements in the cloud, desktop, mobile, and home sectors.

On March 17, 2026, Baidu officially unveiled this product matrix at AI DAY.

Among them, the cloud-based 'Du Claw' provides out-of-the-box web services for users with no prior experience; the desktop 'Du Mate' is a self-developed desktop product; the mobile 'Red Finger Operator' is the world's first mobile lobster application; and the home 'shrimp' is the world's first home lobster after integrating Du Claw into Xiaodu smart speakers.

Compared to the single entry points chosen by ByteDance, Tencent, and Alibaba, Baidu endeavors to cover all conceivable scenarios where users might utilize AI Agents through a 'take-all' strategy.

2. Search Skill Emerges as a Focal Point

Within this strategic framework, Baidu's Search Skill has emerged as the most attention-grabbing offering. On the Open Claw official skill store, Baidu's Search Skill has been downloaded over 45,000 times, becoming the world's most downloaded official search engine skill plugin.

In response, Shen Dou, Baidu Group's Executive Vice President, stated, 'Model training has cycles, and the data internalized by models lags behind. Open Claw requires first-hand, fresh data to solve problems, making search the optimal solution.'

Shen Dou even asserted, 'Search will become a vital infrastructure for the future implementation of agents like Open Claw.'

Search is Baidu's forte. It appears that Baidu is genuinely at the vanguard of a transformative era.

However, there exists a disconnect between skill downloads and commercial monetization. Currently, Baidu's Qianfan platform hosts 1.3 million agents, with over 10 million daily tool calls, but platform activity and commercialization rates remain undisclosed. In contrast, ByteDance's Ark Claw captures official mirror site entry points, Tencent's WeChat Claw Bot leverages 1.2 billion monthly active users, and Alibaba's JVS Book notebook establishes a cloud-to-terminal link. All three strategies aim for ecological spillover after single-point breakthroughs, while Baidu's multi-pronged approach confronts significant implementation challenges.

3. Divergent Strategic Assessments

The crux of this divergence lies in strategic priority assessments. Since agent framework competition adheres to the Matthew effect, once a developer ecosystem takes shape, migration costs become exceedingly high. Therefore, Baidu's rationale is that since the agent entry point has yet to be finalized, full-scenario coverage can maximize user acquisition. However, under resource constraints, competition intensifies: cloud-based Du Claw competes head-on with Alibaba Cloud and Tencent Cloud; desktop Du Mate faces pressure from Windows Copilot and Mac Intelligence's native features; mobile Red Finger Operator must penetrate iOS's closed ecosystem and Android fragmentation.

The outcome of parallel development across four fronts is that no front has established a competitive moat. First, the number of people queuing for the second Lobster Market plummeted from nearly a thousand in the first event to around 20, indicating that the technology's novelty period has waned, exposing the shortcoming of a lack of organic traffic. Second, Baidu lacks a WeChat-level social entry point or a Douyin-level content entry point, compelling it to rely on marketing activities for customer acquisition, exacerbating the risk of an inverted relationship between user CAC (Customer Acquisition Cost) and LTV (Lifetime Value). Even though Baidu was the earliest entrant in the 'Lobster Battle,' it fails to retain users, only to watch as newer competitors surge ahead.

II. Baidu's 'Lobster' Dilemma

1. Business Closed Loop Challenges

As one of the few industry players with full-chain capabilities in chips (Kunlun Core), cloud (Intelligent Cloud), models (ERNIE), and agents (Du Claw), Baidu's end-to-end control is its unique advantage.

However, a complete technology stack does not equate to a viable business closed loop, which has long been a persistent issue for Baidu.

First, the positioning dilemma of the mobile Red Finger Operator is emblematic. Its vision is to achieve task closure across apps, such as hailing rides and ordering food, but technical implementation requires surmounting two barriers. One is Android's system-level permission restrictions and iOS's sandbox mechanism, necessitating negotiations with phone manufacturers for pre-installation or root access. The other is the lack of standardized app interfaces, requiring separate development of adaptation layers for each new app, with maintenance costs growing exponentially with the number of apps.

A further issue is the lack of demand validation. Cheetah CEO Fu Sheng calculates that the daily Token cost of a 24-hour AI team built on Open Claw exceeds $100, far surpassing human labor costs. Although Red Finger Operator attempts to solve complex tasks on mobile devices, whether users are willing to run high-cost digital employees on their phones has not yet been validated through PMF (Product-Market Fit). Moreover, the training costs and data sampling for digital employees also present challenges, and they have not yet been commercialized.

If the 'mobile shrimp' is Baidu's hope, then the 'home shrimp' harbors even greater ambitions.

Huang Rongsheng, General Manager of Baidu's Smart Living Group, described a scenario: when the mistress issues the command, 'We're going on an outing tomorrow; help me arrange it,' Xiaodu and Lobster collaborate to generate a detailed itinerary, sync it to each family member's phone, and proactively set personalized alarms.

It sounds like a dreamy scenario, straight out of a movie, but the barriers to home scenarios are also formidable.

Although Xiaodu's 30 million installed base provides a data entry point, agent-based tasks face reliability, privacy boundary, and ecological openness challenges. The outing planning scenario described by Baidu's Smart Living Group involves data integration across four vertical domains: calendars, maps, dining, and transportation. The current siloed state of platform data makes this scenario's realization dependent on powerful API integration, rather than a single technological breakthrough.

2. Vast Differences in Return on Investment

Baidu seems to have once again fallen into the dilemma of 'excellent technology, many products, but no synergy,' mirroring its experiences in autonomous driving and smart speakers. Apollo leads in technology, but the commercialization of Robotaxi has been sluggish, with even rumors of suspension; Xiaodu speakers have impressive sales, but their profitability is questionable.

Moreover, the gap between technological investment and commercial returns is already evident at the financial level. Baidu's technological investment in AI is substantial, but there is still a considerable journey between technological leadership and commercial success.

Baidu's current product matrix strategy essentially employs resource accumulation to mask positioning ambiguities, competing head-on with strong rivals on every front rather than seeking niches in the ecosystem.

Conclusion:

Baidu's 'Lobster' strategy is a blend of full-stack technological capabilities and commercial positioning ambiguity. The parallel development of four product lines covers all conceivable scenarios for agent implementation, but each line faces blockades from industry giants and cost pressures.

However, the essence of agent framework competition is an efficiency race in productizing model capabilities by cloud vendors. Although Baidu possesses a full-chain toolkit from chips to agents, it lacks the scenario capabilities to integrate these tools into user must-haves. As ByteDance, Tencent, and Alibaba establish footholds in the developer, social, and enterprise markets respectively, Baidu's full-stack advantage becomes a quagmire of resource dispersion.

These issues cannot be resolved by Open Claw alone but are the cumulative result of Baidu's long-standing strategic vacillation and organizational rigidity.

Whether 'Lobster' can become a turning point in Baidu's AI strategy depends not on the speed of technological deployment but on its ability to identify the true core battlefield among the four lines and bear the cost of contracting other fronts. If Baidu continues to pursue a 'take-all' strategy, Lobster will repeat the trajectories of Apollo and Xiaodu.

Otherwise, 'Lobster' may well become just another footnote in Baidu's AI strategic history.