How I run autonomous coding agents from my phone with OpenAI Symphony + Linear | Alessio Fanelli (Kernel Labs)

· Source: Lenny's Newsletter · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Emerging Technologies & Innovation · Depth: Intermediate, extended

Summary

Alessio Finelli, founder of Kernel Labs, demonstrates how he orchestrates autonomous coding agents using OpenAI Symphony and Linear, managing them from a cloud-based Virtual Private Server (VPS) rather than locally. This setup allows him to manage engineering tasks, track token usage for cost analysis (e.g., a task costing 221 million tokens), and review agent-generated code via GitHub pull requests, all remotely. Finelli also showcases a "Pokémon card power buyer" use case, where Codex agents automate tasks like extracting PSA certificate numbers for cards over \$1,000, finding underpriced cards on eBay, and real-time pricing at trade shows. This approach highlights AI's potential to automate highly heterogeneous data tasks, enabling small business creation and improving quality of life by offloading context, as exemplified by personal finance management and email screening.

Key takeaway

For MLOps Engineers or entrepreneurs building agentic workflows, consider adopting cloud-based orchestration like OpenAI Symphony with Linear. This setup allows you to manage agents remotely, track token costs per task, and maintain a full history of agent work, improving debugging and future efficiency. You should also regularly review and refine agent instructions and skills to prevent "model drift" and ensure optimal performance.

Key insights

Autonomous agent orchestration via cloud VPS and structured workflows enhances efficiency and historical task tracking for coding and business operations.

Principles

Method

Integrate OpenAI Symphony with Linear to monitor task boards. Symphony creates Codex workpads, generates plans, and manages code reviews via GitHub PRs, moving tasks through "to do," "in progress," "human review," and "rework" states.

In practice

Topics

Best for: Machine Learning Engineer, AI Engineer, MLOps Engineer, Entrepreneur

Related on AIssential

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Lenny's Newsletter.