Malte Ubl

Malte Ubl is the CTO of Vercel, the company behind Next.js and the Vercel AI SDK. Before Vercel, he spent years at Google as a tech lead on the AMP project and various Search infrastructure efforts. He previously organized JSConfEU in Berlin, one of Europe’s most influential JavaScript conferences, for a decade.

Perspective on AI Agents

Malte argues that agents are a new kind of software, not just a faster way to write the old kind. The economic insight: huge swathes of potential automation were never built because traditional software was too expensive for the tacit-knowledge-heavy edge cases — agents make that work viable, so there will be much more software in the future, not less.

He sees four agent archetypes already deployed successfully today:

1. Agent-as-a-service — customer support and similar roles that benefit from 24/7 operation
2. Compressed research — agents do the research phase of a business process, humans still make the final decision
3. Surface existing information — stitch knowledge that already exists in Slack, recordings, and docs into the systems that need it
4. Eliminate boring work — Vercel’s own support agent has a 90% deflection rate and the team’s job satisfaction improved

He is also publicly bullish that model commoditization is good for application builders, and that Europe is quietly taking a leadership role in AI engineering through projects like Vercel’s AI SDK (Lars Gammel, Berlin), Poe (Austria), and OpenCode.

Key Contributions

• Vercel AI SDK — the open-source TypeScript toolkit for building AI-powered applications, now doing over $10M/week in usage
• Chat SDK — hooks agents to chat apps like Slack, Telegram, WhatsApp
• Just Bash — a bash interpreter written in TypeScript that gives agents sandbox-like environments with nanosecond startup times
• Agent harness architecture critique — argues most current agent harnesses wrongly couple the harness runtime with the code-execution environment