Anthropic’s first developer conference put a clear message in front of roughly 500 attendees in San Francisco on Thursday: the company wants autonomous AI agents to become practical collaborators, not just impressive demos.
While much of the AI industry talks about artificial general intelligence, Anthropic’s stated goal of the year is more immediate. It is trying to deploy a “virtual collaborator” in the form of an autonomous AI agent.
AI agents move to the center of Anthropic’s pitch
Anthropic CEO Dario Amodei opened the event alongside chief product officer Mike Krieger. The discussion quickly moved from what current AI systems can do to what the company expects them to take over next.
Amodei put the shift bluntly in a press briefing: “We're all going to have to contend with the idea that everything you do is eventually going to be done by AI systems,” he said. “This will happen.”
That framing gives Anthropic’s developer push a sharper edge. The company is not simply selling a chatbot or a coding helper. It is presenting Claude and related tools as systems that can handle longer chains of work, with humans increasingly directing, reviewing, and coordinating the output.
Krieger described a workplace where engineers become managers of multiple autonomous agents. In his telling, those agents can handle work that ranges from a simple coding task to complex, full-stack development projects across multiple code bases.
He also pointed to a concrete internal result: “It took our technical onboarding time to get engineers up to speed from two to three weeks to two to three days.”
The promise is productivity, but the question is labor
Anthropic’s public line is that agents help employees with tasks rather than replace them. That distinction matters, because the same examples used to show productivity also raise questions about how much work remains for people as the tools improve.
Krieger asked Amodei when there would be the first billion-dollar company with one human employee. Amodei answered: “2026.” The exchange captured the tension at the event. Anthropic talked about AI agents as tools for workers, while also entertaining a future where a vast amount of company execution could be handled by AI systems.
The company’s leaders have been making similar arguments elsewhere. Cofounder Jack Clark has said he expects people to “manage fleets of AI agents,” while Amodei has said software engineers are necessary for now to guide models.
That “for now” is doing a lot of work. The source described models becoming more capable in areas from coding to creative writing, with redundancies appearing likely as those capabilities expand.
For developers, the near-term change is not just faster code generation. It is a broader change in the job description: less direct production in some areas, more orchestration, review, decision-making, and coordination around AI-generated work.
Claude is already deeply involved in Anthropic’s code
The strongest evidence came from Anthropic’s own engineering workflow. In March, Amodei had said that “90 percent of code” will be written by AI within the next six months. At the developer conference, Krieger gave a current figure for Anthropic itself.
“Something like over 70 percent of [Anthropic’s] pull requests are now Claude code written,” Krieger said.
That does not mean engineers have disappeared from the process. Krieger said they are orchestrating the Claude codebase and attending meetings, adding: “It really becomes apparent how much else is in the software engineering role.”
The implication is straightforward. If AI writes a large share of pull requests, the bottleneck shifts. The human role becomes less about producing every line and more about setting direction, checking quality, integrating changes, and deciding what should be built.
Anthropic’s event treated that shift as a product opportunity. A developer platform built around agents needs to persuade users that autonomous systems can do meaningful work across real projects, while still fitting into human-led organizations.
Biomedical research expands the stakes
Anthropic is also pushing Claude into scientific work. The company is offering up to $20,000 in API credits to researchers in biology and genetics.
Amodei said he is particularly excited about Opus’ ability to aid in cybersecurity and biomedical research. He also said: “I think we're just at the beginning of what we can do with the new generation of model in terms of tasks.”
The same capability gains that make the model useful in research also raise safety concerns. Amodei said in a press briefing: “We have found that the [new] model’s abilities in biology are substantially better.”
According to the source, this has contributed to Claude Opus 4’s Chemical, Biological, Radiological, and Nuclear risk level, making it the highest risk model Anthropic has released to date based on its Responsible Scaling Policy.
That detail is central to Anthropic’s position. The company is trying to argue that more capable systems can be deployed into valuable domains while still being evaluated and controlled carefully.
Growth brings Anthropic into the spotlight
The conference also showed a company becoming more public and more commercially assertive. A spokesperson said Anthropic held the event now because it is finally a big enough company to host one.
The company has doubled in size in the past year to 1,300 employees and is valued at $61.5 billion. That growth changes the context for its safety-first reputation. Anthropic is no longer only presenting itself as a careful actor in AI; it is also presenting itself as a major platform company with developers to court.
During the press briefing, Amodei and Krieger also answered questions about an upcoming compute cluster with Amazon. Amodei said “parts of that cluster are already being used for research.”
On worker displacement, Krieger said: “I don't think you can offload your company strategy to something like that.” The comment drew a boundary around what Anthropic thinks agents can do today: they may execute and accelerate, but company direction remains a human responsibility.
Amodei also commented on a provision in President Trump’s megabill that would ban state-level AI regulation for 10 years. He said: “If you're driving the car, it's one thing to say ‘we don't have to drive with the steering wheel now.’ It's another thing to say ‘we're going to rip out the steering wheel, and we can't put it back in for 10 years,’”
Asked what he thinks about most, Amodei pointed to the race to the bottom, where companies cut safety measures to compete in AI. “The absolute puzzle of running Anthropic is that we somehow have to find a way to do both,” he said, referring to competing while deploying AI safely.
That is the core tension behind Anthropic’s developer day. The company wants AI agents to move from promise to everyday infrastructure. At the same time, its own leaders are warning that the more powerful these systems become, the harder it will be to balance speed, safety, and control.