Organizer WeAreDevelopers drew 15,000 attendees to CityCube Berlin for a developer conference less concerned with AI coding demos than with the messier work of running AI inside software delivery.
The July 8-10 World Congress brought more than 500 speakers and technology leaders from companies including NVIDIA, Amazon, Microsoft, Google Cloud, Atlassian, SAP, Salesforce and IBM. The event ran under the patronage of Germany's Federal Ministry for Digital Transformation and Government Modernisation, and the company bills it as the world's largest gathering for developers, AI builders and technology leaders.
The useful signal from Berlin was mechanical, not magical. If AI agents are going to write, review, test or ship code, teams have to decide how those agents exchange context, how their output is evaluated, what runtime boundaries they get, and who is accountable when generated work looks plausible and breaks something expensive.
From code assistants to delivery machinery
Sead Ahmetović, CEO and co-founder of WeAreDevelopers, described a shift from experiments and prototypes to production deployments with users, failures and operating constraints. His framing matched the program: fewer stage tricks, more discussion of evaluation, autonomous workflow security, infrastructure, team design and human oversight.
Thomas Pamminger, the company's chief product officer and co-founder, said the developer's job changes when agents handle more of the code writing. In his view, engineers become more valuable for defining intent, judging output and designing the handoffs between people and agents.
That is a less comfortable story than the usual productivity pitch. Agent-to-agent workflows need more than a chat box bolted onto an editor. They need repositories that agents can reach reliably, sandboxes for untrusted execution, policies for generated changes, and engineers who can tell the difference between syntactically acceptable code and a system that should be allowed near production.
The roster reflected that shift. Thomas Dohmke, co-founder and CEO of Entire and former GitHub CEO, discussed rebuilding the software development lifecycle for agent-to-agent and agent-to-human collaboration. Atlassian's Taroon Mandhana addressed engineering alignment. NVIDIA CTO Michael Kagan spoke on AI infrastructure, while Amazon CTO Werner Vogels focused on production engineering culture.
Other speakers included Markus Rautert of adidas, Malte Ubl of Vercel, SAP executive board member Thomas Saueressig, Sentry CEO Milin Desai, Salesforce CIO adviser Uli Irnich and IBM automation and AI general manager Neel Sundaresan. Federal Minister Dr. Karsten Wildberger also appeared on the main stage to discuss Europe's digital future.
Launches aimed at agentic software work
The conference also served as a launch venue for tools aimed at the production side of AI development. Four years after showing the first public demo of GitHub Copilot on the same stage, Dohmke returned with Entire, which unveiled a distributed Git network in preview. The system mirrors repositories across regions so AI coding agents can clone and pull code without depending on a single centralized host's latency, outages or rate limits.
Google Cloud announced the public preview of Cloud Run sandboxes, a runtime environment for executing untrusted code and AI agent workloads. Bright Data launched Scraper Studio, an AI-powered system that turns plain-English instructions into web scraping APIs and outputs structured Markdown or JSON, according to Bright Data CPO Ariel Shulman.
The 40,000-square-meter expo hosted more than 200 partners, including NVIDIA, Microsoft, GitHub, Google Cloud, Amazon Web Services, Atlassian, SAP, Salesforce, Twilio, Deutsche Bank, GitLab, Cloudflare and IBM.
The AI software-delivery conference in Berlin is the first stop in a three-continent program. WeAreDevelopers plans a North America event in San Jose, California, from September 23-25, followed by a conference in Bengaluru, India, from November 25-26.
