Shopify CEO Considers Disabling Claude Code
Shopify CEO Tobi Lütke publicly stated that he is considering disabling Claude Code internally at Shopify if it continues to refuse to read AGENTS.md and .agents/skills.
Tobi pointed out that the engineering team may use different Coding Agents such as Codex, Cursor, and Claude Code simultaneously, and more tools are beginning to read a unified AGENTS.md, allowing the team to consolidate project specifications, testing methods, development agreements, and Agent instructions in one place.
However, Claude Code currently relies mainly on its own CLAUDE.md and .claude/skills, leading to different Agents in the same code repository potentially reading two completely different sets of rules, resulting in a "split brain" scenario.
This issue has limited impact on individual developers maintaining multiple sets of files, but for companies like Shopify, where many engineers use AI Coding tools simultaneously, it means needing to synchronize two sets of Agent Contexts long-term; if they become unsynchronized, Agent behavior may deviate.
On the surface, this is a dispute over Markdown files, but behind it lies the question of whether project configurations, Skills, and Contexts should belong exclusively to a particular model company or become a public standard readable by all Agents after Coding Agents enter enterprises.
Shopify has begun to provide answers through its corporate purchasing power: no matter how strong the Claude Code model is, companies are unwilling to maintain parallel engineering specifications for a single tool long-term.
From a market mechanism perspective, corporate purchasing power pressures tool compatibility, shifting funds from proprietary configuration tools to Agents that support unified standards, and event-driven statements elevate discussions on cross-tool standards, benefiting competitors that adopt open configurations while pressuring vendors that insist on proprietary formats.
Source: Public Information
ABAB AI Insight
Tobi Lütke, as Shopify CEO, has long promoted the adoption of internal AI tools and has previously expressed support for the coexistence of multiple Coding Agents. This time, he is using the threat of disabling to pressure Anthropic, maintaining his stance on prioritizing enterprise engineering efficiency, similar to other large tech companies requiring suppliers to adapt to internal standards.
On the capital path, pushing for a unified configuration file standard through public procurement threats is motivated by the desire to reduce maintenance costs of multiple tools operating in parallel, shifting resources from the Claude Code proprietary ecosystem to cross-Agent public norms, forming a closed loop from CEO statements to industry standard negotiations.
This is analogous to the early compatibility disputes between IDE plugins and open-source tools, as well as the history of enterprise software mandating open APIs. The current controversy is in the transition phase from personal developer tools to enterprise-level Agent infrastructure, with the industry's focus shifting from model performance competition to the struggle for dominance over configuration standards.
The structural judgment belongs to the reconstruction of the industrial chain, with the mechanism being that once the coexistence of multiple Agents becomes the norm, proprietary configuration files create coordination costs, forcing enterprises to use their purchasing power to promote public standards, leading the Agent ecosystem from closed formats to interoperable specifications.
ABAB News · Cognitive Laws
Corporate purchasing power is the ultimate driver of unified standards
Proprietary configuration files are the source of "split brain" in the multi-tool era
Compatibility is more decisive for corporate adoption than single model performance