AML Cycle 2 results are approaching: What should agent memory benchmarks reveal?

The Agent Memory Leaderboard's AML (Agent Memory Leaderboard) Coding Memory evaluation is approaching its Cycle 2 results, testing whether coding agents can retrieve, filter and reuse historical engineering experience across bug-fix and feature tasks under both relevant and noisy history. The benchmark poses the question of whether an agent that fixed a bug today can apply that experience to a related task next week, using the example of a team's cache-invalidation fix informing a later batch-update feature without being copied blindly. The leaderboard cites Zed's Agent Metrics as making agent responsiveness visible while arguing project memory is what matters over time.

Agent Memory Leaderboard on X: "Coding agents are getting faster. But speed is only one part of maintaining a real project. The harder question: after fixing a bug today, can an agent use that experience when a related task appears next week? Zed’s Agent Metrics makes responsiveness visible. Proj… / X Agent Memory Leaderboard on X: "Coding agents are getting faster. But speed is only one part of maintaining a real project. The harder question: after fixing a bug today, can an agent use that experience when a related task appears next week? Zed’s Agent Metrics makes responsiveness visible. Project memory asks what happens over time." Coding agents are getting faster. But speed is only one part of maintaining a real project. The harder question: after fixing a bug today, can an agent use that experience when a related task appears next week? Zed’s Agent Metrics makes responsiveness visible. Project memory asks what happens over time. Coding agents are getting faster. But speed is only one part of maintaining a real project. The harder question: after fixing a bug today, can an agent use that experience when a related task appears next week? Zed’s Agent Metrics makes responsiveness visible. Project memory asks what happens over time. Imagine a team fixed a stale-page bug by adding cache invalidation after data updates. Later, an agent builds a batch-update feature. That past fix should be a useful clue — not a patch to copy blindly. The agent still needs to check the current code, version, and constraints. That is the question behind Coding Memory evaluation: Can a system retrieve, filter, and reuse historical engineering experience under both relevant and noisy history? At AML, Coding Memory evaluates this across bug-fix and feature tasks. Finding a similar record is only theShow more