Inside the AI Workflow of Every Six Engineers
Date read: 30th October 2025
Blog link
Key Points
- Interviewed the company Every’s six engineer to see how these use AI
Common Themes
- Many use multiple coding engines to see which one comes up with best result
- Plan day using LLMs, get the LLMs to break the tasks down to atomic steps and then use these steps to get the agent to implement the code.
- One maintains a document of everything he learns from the agent to maintain a long term memory he can use in future prompts.
- One spends afternoomn using new tools and morning using already known ones to ensure the job gets done.
- Special prompt to plan into task into actions for AI agents
- Codex by OpenAI is the most popular coding tool and then Claude to refine details
- Some use LLMs to generate tests and edge cases
- Good idea to check line by line and break down code into smaller taks
- Difficulty: LLMs don’t produce code like rest of the code base