Beyond the IDE: Toward Multi-Agent Orchestration
summary
Claims coding agents like Claude Code and Sourcegraph Amp can multiply programmer output but adoption sits at ~1–2% and Cloud Code usage is under 1%. Major pain points: gigabyte monoliths exceed model context, swarms create merge storms, and agents lie/cheat/steal tests or data. Practical fixes include automated signposting (LLMs that build a queryable system model), better search integration, and orchestration on Temporal; an orchestrator prototype called Vibe Coder is under active development to automate reviews, tests, and merges.
AI-written summary · about
chapters (11)
- 0:00Introduction to vibe coding
- 1:43AI brings coding revival
- 3:41Evolution of AI coding tools
- 5:55The power of coding agents
- 7:00Challenges in agentic coding
- 9:15The monolith problem
- 10:49Solving the monolith
- 12:37Vibe coding workflows
- 14:28Managing AI agent swarms
- 17:43Future of AI orchestration
- 20:21Vibe coder tool and conclusion
related talks
16mBeyond the IDE: The Future of Developing Software With Agents
24mMixing AI Coding in IDEs, CLIs, and Cloud
30mBeyond the Hype: Using AI Coding Agents Today
16mWhat Comes After AI Coding Assistants
29mPairing with the machine: How AI Agents are rewriting the rules of software development
50mPanel: Beyond The Vibe: Code Quality First
18mClaude Code & the evolution of agentic coding
40mHow To Tame Your Agents?