OpenAI published an October 14, 2026 case study describing how Amp uses its models for work beyond writing code. Amp CEO Quinn Slack discusses planning, documentation, page design, marketing and managing other agents. The article is a vendor-published customer account, not an independent comparison of agent products or a new guarantee of what any user will achieve.
The article says Amp chooses models based on what works for its users and that OpenAI models account for most tokens used in the product. Slack describes model improvements as a way to expand what an agent can do before, during and after software development. These statements describe Amp's experience and choices, not evidence that one model is best for every workload.
Amp's Sign in with ChatGPT integration is presented as a way to give users access to more tokens and OpenAI models, including GPT-6 Astra. Slack says usage costs can influence how much work people delegate. The case study does not provide a complete plan-eligibility table, price schedule or unlimited-use promise. Users need the applicable product terms before treating the integration as a fixed cost saving.
The company's own 20-person team uses agents for writing, design and marketing work, according to Slack. He describes documentation and page design as examples where output quality matters. His account is not a controlled productivity study, and it does not remove the need to check factual claims, published copy or changes made by an agent.
The article also describes Puck, a conversational assistant within Amp that can help users start and manage agents or find prior technical conversations. Calling Puck uses GPT-Live-1 for real-time voice interaction, including from a watch. Slack's example of keeping work moving while away from a computer is a proposed use, not a promise of compatibility with every watch or unrestricted background operation.
Amp has been testing Ultrafast mode with GPT-6 Astra, and Slack says faster responses could change how people divide their attention across tasks. That is a reported experience, not a published latency commitment or general-availability announcement. The October 14 case study gives examples of an agent workflow, while access rules, costs and accountable review remain separate questions.