The video expresses the creator’s frustration with Anthropic’s Claude Code due to restrictive usage policies, poor communication, and system prompt limitations that have rendered the paid service unusable for their needs. As a result, they have transitioned to alternative AI tools like Codex and Clerk, praising their transparency and developer-friendly features while bidding farewell to Claude Code.
The video discusses the creator’s growing frustration with Anthropic’s Claude Code, a paid AI service they have been using extensively. Despite paying $200 a month and previously finding Claude Code valuable for tasks like UI work, machine setup, and debugging, recent changes by Anthropic have rendered it practically unusable for their needs. The main issues revolve around how Anthropic has restricted usage, particularly targeting integrations like OpenClaw, by banning mentions of OpenClaw in system prompts and headers, causing API errors and blocking requests. These restrictions have severely limited the creator’s ability to use Claude Code effectively, especially for personal assistant tasks and debugging.
Anthropic’s rationale for these restrictions appears to be related to managing GPU resource usage and preventing abuse from heavy users, especially those using OpenClaw inefficiently. The company aims to prioritize GPU availability for researchers and enterprise customers, but the creator argues that Anthropic’s approach is heavy-handed and poorly executed. Instead of implementing smarter rate limiting or caching improvements, Anthropic chose to outright ban certain use cases, which has alienated power users and developers who rely on Claude Code for legitimate purposes. The creator highlights that while Anthropic has made some token usage improvements, the overall handling of these issues has been disappointing.
A significant point of contention is the opaque and inconsistent communication from Anthropic regarding subscription rules and usage policies. The creator references a post by Matt Pocock, a respected developer who has created a course around Claude Code, expressing frustration over the lack of clarity and delays in getting definitive answers from Anthropic. This confusion extends to how system prompts are handled, with some users suspecting that Anthropic is injecting or modifying system prompts to restrict functionality further. The creator shares a personal experience where Claude Code refused to assist with a Dropbox issue, citing that it was outside its software engineering scope, which they believe is a result of these system prompt changes.
In contrast, the creator praises alternative tools like Codex and Clerk, which offer better developer experiences, transparency, and support. They describe Codex as a superior CLI tool with smarter models and open-source accessibility, making it a preferable replacement for Claude Code. Clerk is highlighted for its ease of integration and comprehensive features for modern applications, including billing and user management. The creator has already shifted their workflow away from Claude Code to these alternatives, signaling a significant change in their AI tool preferences due to Anthropic’s recent missteps.
Ultimately, the video serves as a candid farewell to Claude Code, with the creator expressing disappointment in Anthropic’s direction and a loss of trust in their platform. They emphasize that despite their previous support and willingness to defend Claude Code, the recent restrictions and poor communication have forced them to move on. The creator encourages viewers to explore other AI tools that are more transparent and user-friendly, closing with a hopeful note that future models or companies might better meet their needs. This video marks their final detailed critique of Claude Code, signaling a definitive end to their usage of the service.