AI Teammate Lifecycle Engineering (ATLE) is one of the structured engineering activities proposed in Agentic Software Engineering: Foundational Pillars and a Research Roadmap as part of Structured Agentic Software Engineering (SASE) (SASE), paired there with AI Teammate Infrastructure Engineering (ATIE) (ATIE) under the umbrella of “SE for Agents.” Its stated purpose is to enable agents to retain memory and learn over time, moving them away from “one-off contractors” who start every task from scratch and toward “life-long partners” who retain institutional knowledge.
Usage
The paper grounds ATLE in persistent memory — agents embedded with long-term memory of project history and decision logs, maintaining continuity across tasks without the coach repeatedly supplying the same guidance — citing the DeepWiki used by Devin as an early example of an agent building and referring to its own documentation and decision logs across tasks. It also covers proactive maintenance: an agent with persistent memory and codebase access scheduled during idle compute cycles to scan for technical debt, identify documentation gaps, or propose refactorings, with such proposals filed as new BriefingScripts entering the standard workflow for human review. The paper’s research roadmap for ATLE calls for mechanisms retaining project history, decision logs, and architectural rationale (both continual-learning methods and external memory structures such as graphs, vector stores, and decision records, with SE-specific compression to avoid overflowing context windows), research into scheduling and valuing proactive maintenance work without distracting teams from higher-priority development, and empirical and economic models — rather than anecdotes — of how agentic SE changes the cost model behind long-standing principles (e.g. code duplication becoming easier for agents to update consistently).
Related Terms
AI Teammate Infrastructure Engineering (ATIE), Agent Execution Environment (AEE), Structured Agentic Software Engineering (SASE)