If you just joined me (hi)
Buttfart is a dual-framework where public agents learn from how humans interact with their product
It can also overhaul local agents into evolving personal assistants that learn from how you work, talk, and live
Agents powered by Buttfart OS never stop improving
ButtCore OS:
https://t.co/E2WylDZ3KM
FartOS:
https://t.co/ERHUUKbjmD
Docs:
https://t.co/tmYBcwy0vP
The problem with most frameworks is - they're static, and agents are ever-evolving by design
@NousResearch is doing groundbreaking work with Hermes (adapting to the operator is huge and the project is an inspiration)
The design philosophy is similar, only turned outward - instead of looking at what the operator likes, Buttfart looks at
- What the market liked
- What customers liked
- What socials liked
(or any other type of hard data)
https://t.co/tmYBcwy0vP
Hi, this is a (late) reassurance post
Yes, day 1 is volatile and a little crazy
If Buttfart needs time to establish itself and gain trust, I'm perfectly fine with that and went into it expecting no less
It wouldn't exist without @Solana (hence the name) and I am grateful to the market, volatile or not
Thank you for joining me, I hope you stick around
Notifications on!
Try the framework:
https://t.co/K18wk9EdTc
Docs:
https://t.co/tmYBcwy0vP
To do this, multiple agents with individual roles can now hand off/spawn/report. Every brain evaluates:
- Intent
- Sentiment
- Contradictions
- Relevance/Recall
You decide what data matters the most, agents train themselves on it.
Works for chat, metrics, on-chain, tickets, comments.
https://t.co/K18wk9EdTc
One use-case (running on BFOS right now):
- Buttfart agent manages X account
- Agent receives OHLCV + @Pumpfun token data + xAPI
- Developer sets the exploration tone (experiment a little, experiment a lot)
- Agent posts; agent monitors outcome
- Agent monitors user sentiment, monitors bug reports
- Reports flagged to human
- Improvements deployed to tech
- Permanent change in agent behavior
- Repeat
https://t.co/K18wk9EdTc
Major release now out (fartOS v0.2.0):
✅ Intent is now decided by a model
✅ Each judgment unit can use its own model, role, and prompt
✅ Confidence is now earned by repetition
✅ Losing variant auto-retirement update
✅ Guardrails can now veto a "winning" experiment
✅ ingest(json) now takes any shape - no adapter required
✅ More durable queues
✅ Event processing hotfix
✅ More strict hierarchy: (global → tenant → project → user → session → agent)
✅ Better multitasking
✅ Deterministic test tools added
https://t.co/K18wk9EdTc
One use-case (running on BFOS right now):
- Buttfart agent manages X account
- Agent receives OHLCV + @Pumpfun token data + xAPI
- Developer sets the exploration tone (experiment a little, experiment a lot)
- Agent posts; agent monitors outcome
- Agent monitors user sentiment, monitors bug reports
- Reports flagged to human
- Improvements deployed to tech
- Permanent change in agent behavior
- Repeat
https://t.co/K18wk9EdTc
Buttfart OS can power:
- xAI agents that improve from user engagement
- @Solana dev-agents that improve from how their token performs
- Corporate agents that scope common user problems
Agents running on Buttfart can run research & email their human - or build and deploy themselves.
Get it:
https://t.co/K18wk9EdTc
Buttfart OS allows agents to run and manage entire @solana projects - start to finish - with onchain stats being the success metric.
From the X account - to the tech stack itself.
Agents evolve and improve with every event - and learn to pursue only what works.
Buttfart OS is comprised of two main components.
When equipped with the public-facing framework agents, evolve, accumulate knowledge, and improve, based on how humans interact with them.
- xAI
- Public personas
- Corporate agents
- Support assistants
- Technical project-builders
https://t.co/K18wk9EdTc
Introducing: Buttfart
A dual-framework agentic OS for evolving assistants and systems.
HDqjibDqGg4yDm54TCK8Q4QJVGfAE8Q2d7tygLpZpump
https://t.co/YJ9Mn6zVuN