The Future of Data Teams Is One Person and a Fleet of Agents | Scott Breitenother (Kilo Code) скачать в хорошем качестве

The Future of Data Teams Is One Person and a Fleet of Agents | Scott Breitenother (Kilo Code)

Scott Breitenother built Brooklyn Data Company from his apartment into a 100-person data consultancy, then watched himself become the bottleneck. Now he's co-founding KiloCode — an open source AI coding agent and agentic engineering platform — with Sid Sibrandi (founder of GitLab). His team runs on AI-first principles: one full-stack data engineer plus four autonomous AI agents replacing what used to take a team of five. We go deep on AI coding agents, agentic workflows, AI-assisted code review, and what happens when you let AI agents run autonomously on real engineering workloads. Scott explains why data teams that wait to adopt AI pair programming and LLM-powered tooling will fall behind — and how KiloCode uses parallel agents, CLI agents, and its own VS Code AI extension to ship features at "kilo speed." In this episode we talk about: 00:00 Intro 01:22 From management consulting to Casper to founding Brooklyn Data 03:11 Excel as a gateway drug to the modern data stack 05:30 Why the modern data stack hasn't had a magical moment in years 12:24 The underrated power of asking "I don't know" 19:48 The Slack bookmark hack: how founders stop being the bottleneck 23:49 Data democratization: why AI might finally make it work 28:51 "Exoskeletons, not robots" and what AI actually changes for data teams 32:10 The anti-collaboration philosophy: one human, four agents 42:22 Code is cheap: why Kilo engineers ship on day one 44:50 Data teams can't afford to be two years behind 46:30 Long-running autonomous agentic workloads as the future 48:26 Worst production failure: when an escalator breaks, it just becomes stairs 50:08 Lightning round: Excel, census data, tabs Key topics: AI coding agents vs. traditional development, building AI-native teams from scratch, autonomous agentic workloads, AI-assisted code review, the dbt MCP server for data engineering, vibe coding in production, full-stack data engineer as the new role, when to let AI agents run unsupervised, and the founder bottleneck pattern in startups.