Your ability to write code won't be the thing that earns you money anymore.
Mike Rispoli sits down with Adam Creeger, CTO of Slate and creator of iloom.ai, an orchestration tool for AI agents working with the likes of JIRA and Linear.
Adam's philosophy: sub-agents aren't about playing house. They're about context isolation. Not creating perfect front-end and backend developer personas. But using workflows to avoid context rot and get predictable results.
His approach with swarms? Every agent follows the same workflow: analysis, planning, implementation, code review, verification. Rigorous. Not crazy fast. Not token efficient. But it gives better end results and insulates you from model variance.
The tradeoff? You spend more tokens upfront. But you don't repeat work. And you get an audit trail showing exactly what the agent was thinking and why.
Adam also calls out the industry's broken relationship with issue trackers. Engineers used them for years. Then agentic coding arrived and everyone started littering repos with markdown files. The obvious place for context? The issue tracker.
Topics covered:
- Why Slate is positioned for AI content creation shift
- Sub-agents as context isolation, not role-playing
- Building agent swarms with strict workflows
- Rigorous processes insulate from model variance
- Issue trackers vs markdown files for agent context
- TDD with Claude and better unit tests
- Rails conventions preventing agent chaos
- Quality vs speed tradeoff disappearing with AI
- Why coding ability won't earn you money anymore
- Niche technical areas vs product engineering
- Designing interviews with AI in mind
If you're building with agentic coding or managing teams, this is required listening.