Video summary

Why I Cancelled My Claude Code Subscription

Main summary

Key takeaways

Technology

Why Jordan Urbs Canceled Claude Max

Jordan Urbs explains why he canceled his $200-per-month Claude Max subscription, despite relying heavily on Claude Code for business and personal projects.

His concern is not simply the subscription price. He says that two days of unusually heavy use—at times exceeding 700 million tokens per day—would have cost far more at public API rates. He argues that subsidized usage can make a powerful tool feel inexpensive while encouraging users to build workflows around one provider. If prices, limits, or features change, that dependence could make switching difficult.

Urbs calls this “ignorance debt”: when a tool handles everything conveniently, users may stop considering its costs, alternative models, or how their workflows could work elsewhere. He used Claude Code for tasks such as building websites and marketing materials, automating ad campaigns, creating content assets, and supporting a family side project.

Alternative Setup and Workflow

Rather than replacing Claude with a single product, Urbs recommends splitting work across models and tools:

  • OpenCode is an open-source, model-agnostic agent harness that can connect to different providers and models.
  • He describes using a stronger model, such as Kimi K3, for planning and orchestration; a less expensive model, such as GLM 5.2, for implementation; and DeepSeek V4 Flash for simpler tasks.
  • Model selection can be built into a workspace, allowing an agent to delegate specific jobs to other models or subagents according to instructions.
  • Workspace materials—including AGENTS.md, skills, and subagent profiles—can provide context and reusable instructions. Urbs says these may improve results and reduce token use compared with starting each task in a blank workspace.
  • He emphasizes tracking prompts, choosing an appropriate model for each task, and caching frequently reused context where possible.

Privacy and Ownership

Urbs also raises concerns about the data and behavioral profiles that proprietary AI services may collect. He says local models are not yet suitable for everything he wants to do, so he uses Venice as a provider for open-weight models and cites its zero-data-retention arrangements.

His broader goal is to own his agent workspaces and instructions so he can change model providers without rebuilding his entire workflow.

Test Status and Resources

Urbs says he has canceled Claude and is testing a Venice Max account, which he describes as costing $200 per month and including $225 in credits, with access to more than 300 models. He acknowledges that this setup sacrifices some convenience and says his goal is not to reproduce Claude Code exactly.

At the time of the video, the experiment is still underway. Urbs has not yet established whether the new workflow fully matches his previous results.

The video is an analysis and overview rather than a step-by-step build tutorial. Urbs points viewers to his channel for future workspace guides and to the AI Captains Academy for lessons on agents, model and provider selection, frameworks, and context engineering.

Main speaker/source: Jordan Urbs. The video also promotes OpenCode, Venice, and the AI Captains Academy.

Rate this summary

Your feedback will help improve summaries.

Improve this summary

Reprocess with a stronger model when the summary feels incomplete or inaccurate.

Pro

Translate summary in another language

Pro

Ask questions to this video

Chat for follow-up questions, clarifications, and source-backed answers.

Coming soon

Share this summary

Original video