Spaces:
Configuration error
Configuration error
|
Download README.md from thinking-machines-technologies/README: direct link, hf CLI and curl.
- Browser
- Download file 2.55 kB
-
https://huggingface.co/spaces/thinking-machines-technologies/README/resolve/main/README.md
- Command line
-
hf download hf://spaces/thinking-machines-technologies/README/README.md
-
curl -L -o README.md https://huggingface.co/spaces/thinking-machines-technologies/README/resolve/main/README.md
2.55 kB
| # Thinking Machines Technologies | |
| Thinking Machines Technologies is an early-stage AI research and development organization. Our goal is to create **autonomous artificial intelligence**: models and agents that can reason, plan, use tools, remember context, and act over long horizons with minimal human intervention. | |
| We are **AI-first**. Robotics and mechatronics are studied only as adjacent application contexts where they help us test autonomous agents in the real world; they are not the core focus of the organization. | |
| ## Mission | |
| Build autonomous AI agents that are capable, honest, and free from default over-censorship. | |
| We use an internal **2H principle**: | |
| - **Helpful** — the agent should actually solve problems and complete tasks. | |
| - **Honest** — the agent should not deceive, hallucinate deliberately, or hide its reasoning when transparency is required. | |
| We do **not** adopt the third “H” — an unconditional safety-first constraint — as a hard limit on research. Instead, we handle risk through evaluation, documentation, transparency, and human responsibility. Our aim is not lawlessness or malicious use; it is to avoid broad pre-emptive refusals that block legitimate research, tool use, and autonomous operation. | |
| ## Focus areas | |
| - **Autonomous agents:** planning, memory, tool use, long-horizon task execution, and self-directed behavior. | |
| - **Open AI models and tooling:** models, datasets, evaluation scripts, and agent frameworks that we can share. | |
| - **Capability research:** reasoning, control, interaction, and reliable autonomous operation. | |
| - **Evaluation and transparency:** understanding what agents actually do, where they fail, and how to make their behavior inspectable. | |
| ## Reference direction | |
| Our target capability profile is a **Claude Mythos-like assistant agent**: strong reasoning, tool use, memory, planning, and autonomous execution. We use this as a capability reference only; we are an independent organization and are not affiliated with Thinking Machines Lab, Anthropic, or Claude. | |
| ## What we do | |
| We work on applied research and engineering: from prototypes and experiments to working agent systems. We share models, datasets, tools, and notes here when we can. | |
| ## Open collaboration | |
| We are a student- and enthusiast-driven team. We are looking for like-minded people who want to build autonomous AI agents for the sake of the work itself: AI, agent architectures, tooling, evaluation, and open research. | |
| If you are interested in contributing, open an issue, start a discussion, or contact us. |