DeepMind's Schmid: Build an AI Coworker With 7 Needs Using Gemini Managed Agents
AI Engineer · youtube · 2026-11-06
Philipp Schmid of Google DeepMind demonstrates building an autonomous AI coworker for finance with Gemini and the Gemini API's Managed Agents. He frames a coworker as having seven needs — a brain, a computer, playbooks, tools, credentials, house rules and a schedule — and builds each one:
- Skills, playbooks and files living in the agent's own sandbox
- Built-in tools plus MCP servers for external capabilities
- Credentials swapped in by a network proxy, so a prompt injection can't steal them
- Hooks for house rules requiring human approvals
- Cron triggers and schedules for autonomy
He wraps up with a live demo of an agent that audits a portfolio and updates a Google Sheet on a schedule.
More from coding & agent
- Slack agent Biscuit iterates a party-poll app live, even shipping lazy requests — blixt · 2026-11-06
- Using GrokBot as a 24/7 collector that turns Notion into a live idea database — huangyun_122 · 2026-11-06
- Custom GPTs die Dec 11: if you use someone else's GPT, you can't migrate it — Ok_Negotiation_2587 · 2026-11-06
- ~780 AI slop accounts identified on X; users deploy agents to auto-block them daily — TheZachMueller · 2026-11-06
- Dev runs agentic PR reviews in worktrees, agrees with the agent ~70% of the time — tomaarsen · 2026-11-06
- Formal proofs of tens of millions of lines of code are now feasible, says Anthropic engineer — morqon · 2026-11-06