Shared AI Budgets, Open Agent Systems, and eBay’s Shipping Gap

Podcast


Welcome to P3 Media’s AI Commerce Brief, your daily update on the AI and commerce stories shaping how companies build, sell, and grow. It’s Monday, November 2. Let’s get into it.

OpenAI is changing how paid customers account for agentic work. An updated help document says that, starting in December, Plus, Pro, and Business plans will share one usage allowance across ChatGPT and Codex. Usage will be measured in tokens, while instant chat will not count against that allowance. OpenAI says customers will be able to see remaining usage and reset times in Settings. People who reach a limit can wait for a reset, upgrade, or buy credits for eligible features.

For operators, the important point is that chat and delegated work will draw from one budget. A long coding task can therefore affect the capacity available for other agentic work under the same plan. OpenAI has not published the exact December start date or the allowance size for each plan in that document. Those are the details to watch before teams revise usage policies or budgets.

DeepSeek has released public code for a different approach to coding agents. The DeepSeek Harness team posted a project called dsh-superagent on GitHub on October 31. The repository describes it as a research prototype, separate from DeepSeek’s official products and short of production-ready software.

The prototype focuses on one GitHub repository. It can listen for requests on GitHub and, optionally, Feishu. When an authorized person asks for work, the system can open a work item, give it a separate agent session and directory, and return the result to the original conversation, including through a pull request. A web panel shows items that need human attention.

The broader signal is that global competition is moving beyond model quality. Labs are also competing on how agents receive work, preserve context, and hand results back to teams. DeepSeek warns that the public code may contain bugs and should be reviewed and adapted before use. That makes this an architecture release that still needs work before enterprise use.

In commerce, eBay has expanded the scope of its managed shipping program for selected United States sellers. Its current help page says eligible Buy It Now listings can be priced at $500 or less. The program assigns shipping, provides a prepaid label, updates tracking, and includes USPS, FedEx, or eBay drop-off locations. The new eBay drop-off option is being introduced through local businesses in selected United States cities.

There is an important documentation gap. eBay’s separate shipping terms still say only fashion items priced at $200 or less are eligible, and they cap loss or damage coverage at the lower of the purchase price or $200. For sellers handling higher-value inventory, the practical takeaway is simple: confirm the protection terms before relying on the new $500 listing limit.

Finally, Axios reports that OpenAI is preparing a policy and public-relations push after Tuesday’s United States midterm elections. The reported agenda includes new ideas for safeguards on the most powerful AI systems, community commitments around data centers, proposals intended to support workers and business creation, and a push for international standards. Axios also reports that OpenAI wants to use the phrase AI compute campuses for its infrastructure.

This is a reported plan, and OpenAI has not adopted it as policy. The commercial question is whether OpenAI’s proposals can reduce resistance to data-center development while shaping the rules that govern frontier systems. Watch for the actual documents, the legislative language, and whether communities and lawmakers accept the framing.

That’s your AI Commerce Brief for today. Thanks for listening.

Book A Call