The Pulse
OpenAI Casts Amp as an Agent Platform Beyond Coding
OpenAI’s October 14 profile describes Amp using its models for planning, design, marketing and voice-controlled agent tasks alongside software development. Amp CEO Quinn Slack says some users consume 10 billion tokens a week and that OpenAI
AI.info Team · Written with AI ·
“Our users really appreciate the consistency with which OpenAI models have been getting better and better.”
Quinn Slack, CEO and member of technical staff, Amp
A coding agent takes on the work around code
OpenAI’s October 14 profile of Amp presents the product as more than a coding assistant: an agent intended to help builders decide what to make, create the software and design, then keep working after launch. The account comes from OpenAI’s customer-story page, which describes Amp’s use of OpenAI models and quotes Slack, the company’s CEO and a member of its technical staff.
Amp recommends models based on what it believes works best for its users. OpenAI says its models account for the majority of tokens used on Amp, but the profile gives no market-share figures, independent comparison or breakdown by model. The account positions OpenAI’s offerings as a major part of Amp’s product without quantifying how that usage translates into customer results.
Some Amp users consume 10 billion tokens a week
Slack says some Amp users go through 10 billion tokens weekly. He argues that closely tracking each increment of usage can discourage people from giving agents larger assignments. Amp’s Sign in with ChatGPT integration, he says, gives users access to more tokens and OpenAI models, including GPT-6 Astra.
Slack describes the intended shift as letting an agent work on a longer task without users counting each token. He says customers use Amp to plan projects, choose what to build and continue work after shipping. OpenAI’s profile does not list subscription prices, token limits or the terms of the integration, so it leaves unclear what the added access costs and how much more usage customers receive.
GPT-6 Astra moves into design and marketing
Amp’s own 20-person team uses agents for work beyond software, including marketing, according to OpenAI. Slack says the agent helps improve the experience for new users and supports discovery through search and other channels. GPT-6 Astra also helps the team produce documentation and design pages, areas where Slack says Amp has a distinctive style.
The profile offers a personal example: Slack gave GPT-6 Astra plans for his house, then used the model to create a 3D version in Blender. He and his children turned it into games, including one with a lazy river running through the house. The example shows the range OpenAI wants readers to associate with the model, though the story provides no evaluation of the resulting design or comparison with other tools.
Puck adds voice control and longer-running tasks
Amp’s Puck assistant helps users start and coordinate other agents, or retrieve an earlier conversation about a technical issue. A feature called Calling Puck, powered by GPT-Live-1, lets users speak with it in real time, including from a watch. Slack describes a use for managing agents while away from a computer; OpenAI does not specify when the feature will be broadly available.
Amp is also testing Ultrafast mode with GPT-6 Astra. Slack says faster responses could reduce the need to keep several agent tasks running at once, or leave agents more time to check their own work. The profile gives no speed measurements or test results, and frames Ultrafast as an experiment rather than a generally available capability.
The concrete product direction is broader than code generation: Amp wants agents involved in planning, design, marketing and post-launch work, with Puck as a conversational control point. OpenAI’s account makes that case through Slack’s examples, while leaving the key commercial and performance measures—pricing, model comparisons and Ultrafast benchmarks—unreported.