Google在人工智慧编程工具领域正逐渐落后于OpenAI和Anthropic。尽管去年发布的Gemini 3模型在多项基准测试中超越了主要竞争对手，但Google目前在AI编程方面的表现令人担忧。问题的根源并非技术能力不足，而是公司内部繁琐的官僚体制和管理层层叠叠的红线所造成的阻碍。近几个月来，多位重量级研究人员相继离开Google的AI部门，包括协助发明Transformer架构的Noam Shazeer以及因蛋白质折叠研究获得诺贝尔奖的John Jumper，他们分别投奔OpenAI和Anthropic。

Google在产品开发方面的混乱历史严重影响了其AI编程工具的竞争力。公司内部的晋升机制鼓励管理者推出新产品后迅速转向其他团队，而非持续维护和改进现有产品。这导致Google同时推出了Jules、Gemini Code Assist、Firebase Studio、Antigravity等多个彼此竞争的编程工具，缺乏统一的聚焦和协调。此外，计算资源的分配问题、要求科学家等待六个月确认研究是否适用于Gemini后才能发表论文的政策，以及Mountain View与伦敦DeepMind实验室之间的跨大西洋紧张关系，都进一步加剧了内部的不满情绪。

尽管面临这些挑战，Alphabet公司并未面临生存威胁。该公司拥有自研AI晶片、自有数据中心、健康的云端业务，以及每季度带来巨额收入的线上广告业务——2026年第一季度收入达770亿美元，同比增长15%。然而，Google曾经因未能将自家员工发明的Transformer技术转化为可行产品而错失大型语言模型的先机，如今在AI编程领域可能正在重蹈覆辙。公司已表示将把AI编程工作统一整合到Antigravity旗下，但专栏作者认为，CEO Sundar Pichai需要对公司的激励机制和管理结构进行更根本性的改革，才能在AI竞赛中保持领先地位。

Google is losing ground in AI coding despite its strong technological capabilities. After catching up with OpenAI and Anthropic last year through the release of Gemini 3, the company is now falling behind in AI coding tools—the most lucrative and scientifically important avenue in the field. The issue is not a lack of technical prowess but rather internal bureaucratic dysfunction. Several high-profile departures, including transformer co-inventor Noam Shazeer and Nobel laureate John Jumper, underscore growing dissatisfaction within Google's AI division, as top talent migrates to rivals that offer both stock-option upside and the prestige of working at the frontier.

Google's chaotic product development culture lies at the heart of its AI coding struggles. The company's promotion system rewards launching new products over maintaining existing ones, resulting in a proliferation of competing coding tools—Jules, Gemini Code Assist, Firebase Studio, and Antigravity—with little coordination among them. Compounding this fragmentation are disputes over computing resource allocation, a six-month publication embargo requiring researchers to first check if their work applies to Gemini, and persistent transatlantic tensions between the Mountain View headquarters and DeepMind's London lab led by Demis Hassabis. Former employees have pointed to these systemic issues as a 'systems problem' driven by misaligned incentives across teams.

Despite these challenges, Alphabet is not facing an existential crisis, thanks to its self-designed AI chips, proprietary data centers, robust cloud business, and dominant advertising revenue that generated $77 billion in the first quarter of 2026 alone. However, Google risks repeating its earlier strategic blunder of failing to commercialize the transformer—a technology invented by its own researchers but ultimately capitalized on by OpenAI to build ChatGPT. The company has announced plans to consolidate its AI coding efforts under Antigravity, but the columnist argues that CEO Sundar Pichai must undertake a broader overhaul of the company's incentive structures and management practices to remain competitive in the rapidly evolving AI race.