I wanted to understand how my money moves, so I utilized the JarvisCore framework to help me build a system that does exactly that.
The code is public if you want to pull it apart: github.com/Prescott-Data/off…
Join us for the next build: discord.gg/QxzpRZ6VaD
The structure of Athena memOs and the filters that determine what agents remember long term.
Full session: youtu.be/x_u7rxfhA5U
Discord: discord.gg/jbaZWr3mA
The concept behind how we determine what our agents get to remember for good visa vi what gets discarded. Something you come back to on different days fades slower.
Full session: youtu.be/x_u7rxfhA5U
Come build with us: discord.gg/jbaZWr3mA
I recently yapped about how agents choose what to remember and the part our memoryOs plays in all of this...
Catch the full episode on Youtube: youtu.be/x_u7rxfhA5U?si=NtxZ…
Join us on discord for more agentic convos: discord.gg/jbaZWr3mA
Are you building agents that survive in prod? Check out the JarvisCore framework especially if agent to agent communication and memory are issues you've struggled with when building.
My hope for JarvisCore is that it shows a completely different way to build AI agents. Not as an experiment, but as an approach we've proven.
We open-sourced it a few months ago and kept building. Today it launches on Product Hunt. Here's what makes it different:
• No master agent. Agents are equal peers in a decentralised P2P mesh
• No MCP servers needed
• Zero-trust credentials
• Bounded memory (agents forget on a curve)
If you're building agents, I'd love to hear what's breaking for you.
Picture a crop advisory agent that has talked to one farmer 400 times. At planting time it does not need all 400 conversations.
It needs only three, her soil is acidic, she limed in March, she prefers manure. The rest is history.
Five mentions in one afternoon count once. Five mentions on five different days keep it and thats how an agent determines what to keep or forget when using Athena.
Join us on discord for more like these: discord.gg/YY7MWGywc
In today's office hour I walked through how Athena chooses. It borrows the forgetting curve Hermann Ebbinghaus measured in 1885 and human memory.
Such that memories fade with time, and coming back to them on different days makes them fade slower.
My first instinct used to be to keep more. Longer context, bigger stores. However, everything an agent keeps, it can send to a model again with the next question.
What you keep is what you re-send, and what you must be able to erase.
Back again to chat about agent memory🥳.
Join us on discord to never miss a session: discord.gg/jg8QxAPKZ
Join us as @devmarrie takes us through a technical deep dive of Athena memoryOS and the role it plays in agent's long term memory.
Tuesday 28th Sep 11am EAT : lnkd.in/dx6XFUiN
babeforagents retweeted
a nice debut to founder-led distribution "right of passage" without using "hey guys"
okay folks we've open-sourced an agent runtime and we think you will like it for 5 things:
no MCP wiring, no master agent just pure mesh with peer-to-peer, deterministic control, zero-trust through an agent broker, and a bounded memory lifecycle.
babeforagents retweeted
Built for ever evolving and learning AI agents 🤖
How are developers building agent memory on #JarvisCore?@prescottdata 's Marriane takes devs through a build-along session on a garage desk use case during last week's OH!
A mechanic walks up to a vehicle, and before he touches it the agent tells him what this garage has already done to that car, and what got left undone.The first brief opened with the registration and the model. Accurate, and useless cause the mechanic was standing next to the car
The full session available on YouTube: youtu.be/EaICaH-kYZM?si=kj2Y…
Join us on discord for more fun like this: discord.gg/SxTqUv4M8
So I corrected it once, in one sentence. "Tell me who worked on it and what is left, not about the car". It answered "Noted. I will lead with that from now on." Every brief after that opened differently.
The full session available on YouTube: lnkd.in/drAXWDWy
Join us on discord for more fun like this: lnkd.in/dyRkR-x9
That is a memory design decision, not a model capability. It is also why recall has to be governed rather than simply accumulated. Everything this agent remembered stayed inside the boundary, on the machine in front of me.
Here is the part worth sitting with. It always knew about the perished radiator hoses. They were in the first brief, in the second sentence. It did not learn a fact. It learned what mattered specifically to the mechanic.
The agent learnt in real time the expectations of a mechanic at a garage store and helped him better understand the car he is fixing, for more information checkout our discord discord.gg/z69QKEnjd @prescottdevs
See you on Thursday as we tackle why your agent's perfomance might be due to what it actually remembers and not an issue with its logic😌
Join us on the 03 of September for our second build along session, as we build agents that continuously get better from scratch with @devmarrie . Link: discord.gg/jg8QxAPKZ?event=1…
