AI Agent Hermes Tested for Memory Retention Over Seven-Day Continuous Run
agents
| Source: Dev.to | Original article
AI agent Hermes was run continuously for 7 days, revealing significant memory loss.
As we reported on May 17, Hermes Agent has been making waves in the AI community with its ability to run AI models locally. Now, a new experiment has shed light on a critical issue plaguing AI agents: memory retention between sessions. A user ran Hermes Agent continuously for 7 days on the same task, revealing that the skill file on day 7 looked drastically different from day 1, indicating significant memory loss.
This matters because AI agents' inability to retain memory between sessions can lead to a lack of trust from users, as noted in a recent article on Memory Serialization. Hermes Agent's approach to memory, where it's injected at session start and always present, is an attempt to address this issue. However, the experiment suggests that even with this approach, long-term memory retention remains a challenge.
What's next is to watch how the developers of Hermes Agent and other AI models respond to this issue. Will they prioritize improving memory retention, and if so, how? The AI community is eagerly awaiting solutions to this problem, as it's crucial for building trust in AI agents. As Matt Mullenweg noted, trusting remote AI agents can be strange, and local solutions like Hermes Agent may hold the key to overcoming this hurdle.
Sources
Back to AIPULSEN