AI news ·
Amp expands what AI agents can do with OpenAI models including GPT-6 Astra and GPT-Live-1
Amp now runs most tokens through OpenAI's GPT-6 Astra, with some users consuming 10 billion tokens a week. The 20-person team offloads planning, design, and marketing to agents instead of scaling headcount.
Amp, a startup building a frontier AI agent for developers, now runs the majority of its tokens through OpenAI models. CEO Quinn Slack said the company uses GPT-6 Astra, Ultrafast mode, and GPT-Live-1 to push what agents can do beyond code generation-into planning, design, marketing, and real-time voice interaction. The shift matters because it signals a working model for how small teams can offload entire workflows to AI without scaling headcount.
Moving past per-token hesitation
Some Amp users consume 10 billion tokens a week. Slack believes that watching every increment of usage can hold teams back from discovering what an agent can actually accomplish. To reduce that friction, Amp launched its Sign in with ChatGPT integration, which gives users access to more tokens and to models like GPT-6 Astra. "Sign in with ChatGPT means that people can have that feeling where they're not counting every single token," he said. "And once you actually let the agent go and have a longer leash, then you realize just how much it can do."
The integration changes how builders approach larger tasks. Users now let agents handle planning, decide what to build next, and continue working after a product ships-work that previously required constant human direction.
Design and writing that meet a picky team's standards
Amp's 20-person team uses its own agent for marketing, documentation, and user onboarding. "We don't have someone devoted to it, but we have an agent devoted to it," Slack said. GPT-6 Astra produces documentation and designs pages that match Amp's specific visual and editorial style. "We have a very unique design style and writing style. We're very picky about this, but we've been really happy with how GPT-6 Astra does," he said.
Slack has also tested the model on personal projects. He fed it the plans for his house and used it to create a 3D model in Blender. He and his children then turned the model into games-adding a lazy river through the house or chasing garlic around it, a nod to their grandmother's dislike of garlic. "And that was not possible just a couple months ago," he said.
Voice agents and faster workflows
Amp recently introduced Puck, a conversational assistant that helps users launch and manage other agents. Through Calling Puck, powered by GPT-Live-1, users can talk to Puck in real time-including from a watch. Slack sees it as a way to keep agents working while someone is away from their computer, even during a run.
Ultrafast mode with GPT-6 Astra is also changing behavior. Today, people often juggle several agent tasks while waiting for each to finish. Faster responses mean users can run fewer concurrent agents or give an agent more time to check its own work before returning a result. "People have not seen a model that's this good, that's this fast," Slack said.
Why this matters for product and development professionals
Amp's approach shows a practical shift: the agent is becoming the default operator for tasks that span the full product lifecycle, not just code generation. For teams building software, the combination of faster models, voice interaction, and token access through Sign in with ChatGPT removes friction that previously made agent-driven workflows feel expensive or slow. The takeaway is not that agents will replace developers, but that letting an agent own more of the pre- and post-code work-planning, design, marketing copy-can change how a small team allocates its human attention. Professionals looking to build that skillset can explore AI Agent Courses or structured AI for Software Developers Courses to apply these patterns directly.