Running AI Engineer with AI — swyx

Key Concepts

• Tiny Teams: Organizations with high revenue-to-employee ratios, capable of managing large-scale operations with minimal staff.
• Coding Agents (e.g., Devin, Cognition): AI tools that autonomously handle software development tasks, dependency management, and web-based workflows.
• Yak Shaving: The process of dealing with a series of small, tedious, and often unrelated technical tasks required to complete a larger project.
• Agent Experience (AX): A shift in software design where the primary user of an application is an AI agent rather than a human, prioritizing APIs, CLIs, and machine-readable interfaces over traditional UIs.
• Model Context Protocol (MCP): A standard for connecting AI assistants to systems, data, and tools.
• "AGI Pill": A metaphorical term used by the speaker to describe the adoption and integration of AI agents into daily professional workflows.

1. Main Topics and Key Points

Swyx, co-founder of AI Engineer conferences, details how his team of nine people manages a multi-million dollar business and large-scale conferences using AI agents.

• Productivity Evolution: The speaker tracks his journey from focusing on general AI productivity (Year 1) to cost-curve optimization (Year 2) and finally to "Tiny Teams" (Year 3).
• The Shift to Agents: The core argument is that agents are no longer just for coding; they are for "everything else," including data management, research, and operational tasks.
• Parallelism and Autonomy: Agents allow for non-linear workflows where team members can initiate tasks, go to sleep, and have an agent complete the work, effectively removing human bottlenecks.

2. Real-World Applications

• Conference Management: The team uses agents to manage 130+ speakers and thousands of attendees. By treating the conference schedule as code rather than a traditional CMS (Content Management System), they allow agents to handle updates via email or screenshots.
• Procurement and Research: The speaker used an agent to research and source a physical prop (a lobster) for the conference, demonstrating that agents can handle real-world logistics and procurement.
• Design-to-Code: By hooking agents into Figma, the team turned design files into pixel-perfect, functional websites without manual coding.

3. Methodologies and Frameworks

• The "Agent-First" Workflow:
  1. Identify the "Yak Shave": Use agents to handle dependency installation and environment setup.
  2. Bridge the Gap: Connect design tools (Figma) directly to agents.
  3. Asynchronous Collaboration: Utilize agents to process tasks across different time zones (e.g., a designer in Indonesia prompting an agent while the founder sleeps).
  4. Replace SAS: Systematically identify top concerns/pain points of replacing a SaaS tool with a custom-built, agent-managed solution.

4. Key Arguments

• Agents Increase Human Satisfaction: The speaker argues that agents increase employee productivity not just through speed, but by removing the "drudgery" of work, allowing employees to focus on creative tasks (e.g., adding Easter eggs or animations) that they previously wouldn't have had time for.
• The Death of the Dashboard: As agents become the primary users of software, traditional UI/dashboards are becoming less relevant. Developers should prioritize building robust APIs and CLIs.
• Psychosis as a Driver: The speaker suggests that "psychosis"—the willingness to build custom solutions and replace established SaaS tools—is a necessary trait for leaders looking to leverage AI to its full potential.

5. Notable Quotes

• "Anytime there's like random yak shaving, I think one underappreciated benefit of agents is that they save you the yak shaves."
• "I'm no longer talking about agents for coding or like how many lines of code I'm producing. I'm getting more productivity out of my humans."
• "Your custom UI is kind of going away; you should ship UI to somebody else's app."

6. Data and Research Findings

• Scale: A team of nine people manages a business generating over $9 million in revenue.
• Efficiency: The speaker noted a massive spike in productivity, citing "207 replies" in an agent-based workflow as an example of the increased volume of work being processed.
• Industry Trend: Citing Malte (Vercel), the speaker noted that 60% of the Vercel user base is now comprised of bots/agents, not humans.

7. Synthesis and Conclusion

The primary takeaway is that the "future of work" is defined by the transition from human-centric software interaction to agent-centric workflows. By treating agents as team members capable of handling routine, boring, or complex technical tasks, small teams can achieve the output of much larger organizations. The speaker urges professionals to "wake up" to this shift, stop relying on traditional SaaS dashboards, and start building for the "Agent Experience" (AX).