Yohei Nakajima Unveils ActiveGraph Event-Sourced Runtime For Durable AI Agents
Users are excited about Yohei Nakajima's ActiveGraph event-sourced runtime for durable AI agents because its log-first design solves state recovery and replayability issues in long-running tasks.
No Digg Deeper questions have been answered for this story yet.
Most Activity
🆕 ActiveGraph: The Log is the Agent my talk from AI Engineer is live!!! 😆 https://www.youtube.com/watch?v=khVX_BUnEwU it's about @activegraphai, an event-sourced graph runtime for building durable long-running agents: - what it is - code examples - experiments - benefits & surprises please check it out, share with your friends who like to tinker at the frontier, and let me know if you have any Qs!
@yoheinakajima @ActiveGraphAI the minutia of pokemon elo is wild, I really couldn't keep up after breeding became a thing in the games. i've been watching the chronicles of this codex playing through pokemon red and seeing its ballsy choices.
@yoheinakajima https://github.com/escoffier-labs/brigade if you wanna see what it inspired
@blennon_ @ActiveGraphAI “I want to try activegraph, look it up and think of some projects I would like”
@solomonneas @ActiveGraphAI haha after trying a whole bunch of stuff that made incremental improvements, we tried copying an example from the discussion board that outperformed all of the stuff we had tried. and the stuff we tried before didn't improve the example. so i got sad and kinda moved on :)
TL;DR 1 - Instead of building agents around the LLM, ActiveGraph builds everything around the log. 2 - Every action and state change gets recorded, which turns the agent’s state into a graph. 3 - This lets you replay, rollback, or fork the agent when things go wrong. 4 - It’s meant to make long-running agents more reliable and easier to understand. Looks interesting, especially if you’ve struggled with agent state and debugging. Worth watching the full talk. Thanks for sharing.
@yoheinakajima @swyx @ActiveGraphAI Adding this to the ever-growing “must watch later” list. 😂
loved it! how'd the pokemon ai tourny go? this is really helpful for me, I've had a hard time breaking down my similar project to simple &easy to visually understand. this was informative but very casual thanks for sharing, i'll send some updates to you on my project soon also its really cool you got to speak at ai engineer over a personal project, that's so awesome man
just watched @yoheinakajima's Active Graph talk at AI Engineer. not a code call graph but the agent run itself is the graph: goals, tasks, evidence, decisions, all projected from an append-only event log. replay, fork, diff any point. auditability is structural, not bolted on after. go watch Active Graph helped shape how Brigade and its code intelligence (the artist formerly known as graphtrail) functions
🆕 ActiveGraph: The Log is the Agent my talk from AI Engineer is live!!! 😆 https://www.youtube.com/watch?v=khVX_BUnEwU it's about @activegraphai, an event-sourced graph runtime for building durable long-running agents: - what it is - code examples - experiments - benefits & surprises please check it out, share with your friends who like to tinker at the frontier, and let me know if you have any Qs!
@yoheinakajima @ActiveGraphAI i'm super excited to start experimenting with this
@ajmal_jalal26 @ActiveGraphAI good summary :)
@yoheinakajima @ActiveGraphAI Long-running agent tasks are where this becomes very real. We have had to add controls around boundaries, logs, checkpoints, approval gates, timeouts, and resumability. Otherwise the agent can look productive while becoming hard to inspect, steer, or recover.
@yoheinakajima @ActiveGraphAI Will we ever see Mr.Blu again?
@yoheinakajima @ActiveGraphAI 4J9h6PZq4V6Mkg8YUdSMAmh33Vi74zgSBZ3UqGZSpump https://dexscreener.com/solana/HbjQWfpTurGKKbwG9hV3ZRCGW78VX81DK5GyyjUFXBnQ
@yoheinakajima @blennon_ @ActiveGraphAI BK7r5UBN7HxkmqrFgLhJrHfoyJUDWGFA29CA18L7pump 💯 fees send Yohei $ActiveGraph
@buildwbhoomika @swyx @ActiveGraphAI only 15 min!