Home > Videos

🚀🤖⚙️ L8 Principal’s Agentic Engineering Workflow

🤖 AI Summary

• 💻 Master your workflow using a terminal-centric environment like Westerm combined with Tmux for session management and persistence.
• ⌨️ Utilize NeoVim as a text editor to maintain high speed and keep hands on the keyboard, minimizing context switching.
• 🧠 Implement onboard memory using Markdown files that store persistent preferences and project-specific contexts, preventing agents from making repetitive errors.
• 🛠️ Employ skills for progressive disclosure, loading only necessary instructions to keep system prompts concise and token-efficient.
• 🗣️ Adopt voice input for prompts, which is significantly faster than typing, using local transcription tools like Open Super Whisper to maintain privacy and quality.
• 🚀 Design and prioritize tools for agent ergonomics, ensuring they are optimized for AI interaction rather than just human UI, reducing token usage and latency.
• 🎨 Use artifacts and visual editors like Lavish to plan complex work, allowing for easy annotation and feedback that surpasses text-based interactions.
• 🤖 Shift from a code reviewer to an engineering manager role, utilizing pipelines like No Mistakes to automate validation, testing, and PR creation.
• 🌙 Leverage background agents for long-running, repetitive, or verifiable tasks with tools like Good Night Have Fun.
• 🌳 Scale parallel development using Git work trees managed by utilities like Treehouse to prevent conflicts between simultaneous agent sessions.
• 🚢 Recruit a first mate agent to orchestrate multiple crew members, manage overhead, and allow the captain to focus on high-level strategic planning.

❓ Frequently Asked Questions (FAQ)

🧩 Q: How can agentic workflows avoid high token costs while maintaining performance?

🧩 A: Efficient design involves utilizing token-optimized output formats, implementing progressive disclosure of skills so agents only load necessary context, and choosing tools specifically architected for agent ergonomics rather than traditional graphical interfaces.

⚓ Q: Why is a terminal-centric approach recommended for managing AI agents?

⚓ A: Terminal-based workflows enable hands-on-keyboard discipline which reduces context switching, ensures consistent environments across different hardware, and supports powerful multiplexers like Tmux that allow for persistent, parallelized agent sessions.

🛡️ Q: What risks are associated with installing third-party agent skills from the internet?

🛡️ A: Unvetted skills can execute arbitrary code on a machine, potentially leading to the leakage of sensitive API keys or credentials, and many popular skills are not rigorously evaluated, often degrading agent performance instead of improving it.

📈 Q: How does the No Mistakes pipeline improve code quality and developer velocity?

📈 A: The pipeline automates the entire lifecycle of a code change - including branch creation, rebase, adversarial review, end-to-end testing, and documentation updates - ensuring that only verified, high-confidence changes reach the pull request stage, thus removing the human bottleneck of manual diff review.

📚 Book Recommendations

↔️ Similar

• 🧠 Working with AI by Thomas H. Davenport and Steven M. Miller explores how professionals can effectively collaborate with artificial intelligence to enhance productivity and decision-making.
• 🛠️ The Pragmatic Programmer by Andrew Hunt and David Thomas provides foundational principles for software engineering excellence that remain highly relevant when scaling individual output through automation.

🆚 Contrasting

• 📉 Deep Work by Cal Newport argues against the fragmentation caused by constant digital distractions, advocating for intense, focused concentration that contrasts with the multi-tasking agent management described.
• 🎨 The Design of Everyday Things by Don Norman emphasizes human-centered design for physical and digital products, contrasting with the agent-centered ergonomics required for automated engineering workflows.

🎨 Creatively Related

• 🚀 The Phoenix Project by Gene Kim, Kevin Behr, and George Spafford uses a narrative format to explain the critical importance of streamlining workflows and managing bottlenecks in IT operations.
• ⛵ The Goal by Eliyahu M. Goldratt introduces the theory of constraints, a management philosophy that perfectly mirrors the challenge of identifying and optimizing bottlenecks when scaling engineering output.