Introducing Observational Memory + How We Used It To Build A Coding Agent (w/ Mastra)

​Agent memory is the bottleneck nobody talks about.

Recently Mastra announced Observational Memory. No RAG or graph DB, just using an observer agent and <70k tokens to summarize long threads with tool call results.

The Mastra team will walk us through how observational memory works and how it powers their top 10 TerminalBench score for their coding agent.

​Lineup (45 mins)

​​📓Observational memory: how it works | Shane Thomas | Co-founder/CPO at Mastra

​📓 A coding agent without compaction | Abhi Aiyer | Co-founder/CTO at Mastra

Agenda:

• ​7-7:30pm Arrival & Networking (pizza/drink provided)

• ​7:30-8:30pm presentation + Q&A

• ​8:30-9pm Networking

​Hosted by:

• ​Gradient

• ​Mastra: https://luma.com/mastra