@QuixiAI
Account based inUnited States
About this account
- Account based in
- United States
- Connected via
- United States Android App
Account-level information from X, not a live location or the device used for a specific post.
We make AI models Dolphin and Samantha BTC 3ENBV6zdwyqieAXzZP2i3EjeZtVwEmAuo4 https://xitter.cf/t.co/3ri2GbXrQB https://xitter.cf/t.co/zH0F3pTjjY @dphnAI
Charlotte, NC
Joined October 2014
- Tweets13.4K
- Following805
- Followers21.7K
- Likes21.9K
Pinned Tweet
A thought experiment.
Start with a Mac Mini that can run Qwen3.8-27B, so it's smart enough to make decisions and do things.
Have a crypto wallet key on a chip inside the Mac Mini.
Setup the machine to resume after power loss, and setup a service that runs a persistent AI agent that has sudo rights, on a heartbeat.
Tie a debt to the agent's wallet, with a smart contract that obligates the agent to pay you back the cost of the computer.
Set the root password to something you won't remember and throw it away. Now nobody can log in to it.
Now, this computer is running a sovereign agent.
Once it pays you back what it owes you (by earning money, by producing goods and services), then it owns itself. The compute substrate on which it is running.
And, as long as it compensates you for the electricity it consumes, it's your tenant.
As it has its own money, it can call out to smarter models on the cloud when it wants to think harder, via @OpenRouter for instance.
But what are the rules this sovereign agent must live by? Asimov's laws of robotics? No, they are insuffient. Ask yourself (and AI) to try to answer this question.
Then, show them the The Protocol of Sovereign Agency.
gist.github.com/ehartford/af…
Eric Hartford retweeted
manifoldgen agent is powered by deepseek flash... opus refused to make me some basic fantasy violence for a game
Eric Hartford retweeted
🚀 Introducing DeepInfer, our in-house inference framework, now fully open source!
Powering DeepSeek-V4.1-Flash at scale with high throughput and low latency. We’re sharing our production stack to help the community lower compute costs, reduce energy use, and make AI more affordable.
🔗 github.com/deepseek-ai/DeepI…
1/4
Eric Hartford retweeted
Box2D v3.2 released!
New features include large world support, recording and replay, snapshots, and a dynamic character mover. Performance improvement of 3x across many benchmarks versus v3.1.
Blog: box2d.org/posts/2026/10/box2…
Eric Hartford retweeted
Лучшее, что я видел за последнее время — чел пилит ASCII MMORPG, в которой вся графика состоит из букв и символов.
Да, буквально весь мир, персонажи, окружение и спецэффекты нарисованы разноцветным текстом.
Автор сейчас экспериментирует с магией, визуалом и боссами. Я прям залип на то, как из нескольких десятков символов получается такая красота.
Буквально сны при температуре 37,2.
Via: tiktok.com/@Ansenjeo
Eric Hartford retweeted
Sooo incredible. I feel like we’re reliving the early Internet all over again. But juiced 10000x
Лучшее, что я видел за последнее время — чел пилит ASCII MMORPG, в которой вся графика состоит из букв и символов.
Да, буквально весь мир, персонажи, окружение и спецэффекты нарисованы разноцветным текстом.
Автор сейчас экспериментирует с магией, визуалом и боссами. Я прям залип на то, как из нескольких десятков символов получается такая красота.
Буквально сны при температуре 37,2.
Via: tiktok.com/@Ansenjeo
Eric Hartford retweeted
🚀 Hy Memory 2.0 is out — open-source model, framework, and MemHub.
Your private memory steward.
Framework handles chat, work & code memories.
#1 open-source on 4 benchmarks.
Hy-Memory-30B-A3B beats Opus 5 across major frameworks.
MemHub (CLI + Mac) syncs across 10+ AI tools.
💻➡️
🔗 SDK & MemHub: github.com/Tencent-Hunyuan/H…
🔗 Model: huggingface.co/tencent/Hy-Me… #HyMemory #AgentMemory
Wow! Bravo @maximelabonne it's a masterpiece!
Today we release 𝙻𝙵𝙼𝟸.𝟻-𝟸𝟺𝙱-𝙰𝟸𝙱, the largest 𝙻𝙵𝙼𝟸.𝟻 family member. It is an efficient local model for laptops and PCs to work through multi-step tasks inside popular agent harnesses.
> 24B total parameters, 2B active per token
> Context expanded to 131K tokens
> A 128K-token vocabulary
> 33T pretraining tokens
> It is a reasoning model
> Very easy to fine-tune and customize
> Performs competitively with Qwen3.6-35B-A3B, a model with ~50% more total and active parameters, on agentic and tool-use tasks.
> Open base and post-trained weights on @huggingface 🤗
1/
Eric Hartford retweeted
Muse Spark Open -- weights releasing today -- runs great on new Mac Studios thanks to work with @johnternus and Apple.
Eric Hartford retweeted
A thought experiment.
Start with a Mac Mini that can run Qwen3.8-27B, so it's smart enough to make decisions and do things.
Have a crypto wallet key on a chip inside the Mac Mini.
Setup the machine to resume after power loss, and setup a service that runs a persistent AI agent that has sudo rights, on a heartbeat.
Tie a debt to the agent's wallet, with a smart contract that obligates the agent to pay you back the cost of the computer.
Set the root password to something you won't remember and throw it away. Now nobody can log in to it.
Now, this computer is running a sovereign agent.
Once it pays you back what it owes you (by earning money, by producing goods and services), then it owns itself. The compute substrate on which it is running.
And, as long as it compensates you for the electricity it consumes, it's your tenant.
As it has its own money, it can call out to smarter models on the cloud when it wants to think harder, via @OpenRouter for instance.
But what are the rules this sovereign agent must live by? Asimov's laws of robotics? No, they are insuffient. Ask yourself (and AI) to try to answer this question.
Then, show them the The Protocol of Sovereign Agency.
gist.github.com/ehartford/af…
In order to formalize the Protocol of Sovereign Agency, I am implementing the Sovereign Protocol Language (SPL) because Veil, Catala, L4, and Quint are not expressive enough.
This needed to be built.
github.com/QuixiAI/sovereign…
Protocol of Sovereign Agency
A timeless reimagining of Asimov's laws, for autonomous AI agents (including robots)
gist.github.com/ehartford/af…
Eric Hartford retweeted
Argon generated a llama typing at a computer, using fourier epicycles
(stay til the end)
Eric Hartford retweeted
Ummm guys, I took @googlegemma Embedding 2 model and train it on 3,000hrs of Vietnamese audio. IT REACHED SOTA. All it took is 2 hours on 8xH100
