AI agents have an amnesia problem.
They can code.
They can use tools.
They can run in Telegram.
They can live in OpenClaw, Hermes, and local models.
But every new session still starts from zero.
So I built 8mem.
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.
AI will transform every industry, power every company, and be built by every country.
Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.
The world needs both frontier closed models and frontier open models.
https://t.co/AUKzoQ5Ikb
Agree @sharbel Big day - I launched 8mem.
Hermes <8mem> Openclaw
This has been few weeks of intense build/test/fix work. The idea is simple: AI agents should not forget you every time you switch tools.
It now works with OpenClaw, Hermes, Telegram, browser UI, and local Ollama models.
It is 100% free and open source.
Anyone building agents/local AI
GitHub: https://t.co/q50ki2cGlk
@durov Did you guys sneak a look at our dev roadmap? 😉 We literally custom-engineered a rich UI box layout for https://t.co/Bz6kEGGxMI, basic markdown just wasn't cutting it for deep AI comparisons, Native tables,nested lists are going to completely change the game for AI agent UX.
@Teknium@durov No shot you guys just dropped this. Did we accidentally leak our roadmap to the Nous team? 😂 We’ve been locked in custom-engineering a rich UI box layout for https://t.co/Bz6kEGGxMI ,to handle deep AI comparisons. Here is the future of agent UX. Hermes is moving at lightspeed.
@durov Did you guys sneak a look at our dev roadmap? 😉 We literally custom-engineered a rich UI box layout for https://t.co/Bz6kEGGxMI, basic markdown just wasn't cutting it for deep AI comparisons, Native tables,nested lists are going to completely change the game for AI agent UX.
@steipete True, 'do whatever' is a recipe for disaster. Are you using hard-coded schema validations, or relying entirely on LLM-level negative constraints to block unauthorized purchases?
8mem also integrates with Hermes, so I’d especially value feedback from the @NousResearch / Hermes community.
The core question:
How should long-running agents remember, correct, and forget across sessions?
AI agents have an amnesia problem.
They can code.
They can use tools.
They can run in Telegram.
They can live in OpenClaw, Hermes, and local models.
But every new session still starts from zero.
So I built 8mem.
This is the part that matters:
I save a memory in one agent.
Then a different agent, running on a different runtime, can use it.
One memory.
Two agents.
Saved in Viri.
Available in Govi.
GitHub:
https://t.co/Jf8f7j5YhZ
Website:
https://t.co/9QzkZH76rd
The demo connects an OpenClaw agent and a Hermes agent through one shared memory layer.
Would love feedback from @steipete.since 8mem integrates with @openclaw .
The trust features are the real product:
Save a memory -> Saved.
Duplicate -> Already in memory.
Correct it -> Updated.
Forget it -> Forgotten.
Memory is only useful if users can inspect it, fix it, and delete it.
@NousResearch Huge update! Getting media/voice support and interactive buttons natively via hermes whatsapp-cloud is a game changer for building production-grade workflows. Can't wait to test this out