Havenlon/ Execution-Boundary-Qwen38-27B-Q4_K_M

45 Downloads Updated 5 days ago

A Havenlon-focused Qwen3.8 27B model for execution-boundary reasoning, Adversarial Completeness, independent veto, evidence chains, and AI Agent execution control.

ollama run Havenlon/Execution-Boundary-Qwen38-27B-Q4_K_M

curl http://localhost:11434/api/chat \
  -d '{
    "model": "Havenlon/Execution-Boundary-Qwen38-27B-Q4_K_M",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='Havenlon/Execution-Boundary-Qwen38-27B-Q4_K_M',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'Havenlon/Execution-Boundary-Qwen38-27B-Q4_K_M',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Models

Name

1 model

Size / Usage

Context

Input

Execution-Boundary-Qwen38-27B-Q4_K_M:latest

17GB · 256K context window · Text · 5 days ago

Execution-Boundary-Qwen38-27B-Q4_K_M:latest

17GB

256K

Text

Readme

Execution-Boundary-Qwen38-27B-Q4_K_M

A Havenlon-focused Qwen3.8 27B model for reasoning about execution boundaries, Adversarial Completeness, independent veto, evidence chains, and the control of AI-driven real-world execution.

Focus

Execution Control
Execution Boundaries
Adversarial Completeness
Independent veto
Intent-to-Execution consistency
Evidence chains
Hardware-backed execution control
AI Agent execution risk
High-risk automated systems
Fail-Secure behavior

Run

ollama run Havenlon/Execution-Boundary-Qwen38-27B-Q4_K_M

Quantization

Q4_K_M

Core Principle

Havenlon assumes that software systems, AI Agents, APIs, accounts, policies, and approval workflows may eventually be misused, compromised, or operate on incorrect information.

Security therefore cannot depend only on preventing compromise.

It must also limit what a compromised component can ultimately cause.

A local failure should not automatically become a final real-world action.

The execution boundary exists between:

what software wants to happen

and

what reality is allowed to become

Its most important capability is not producing another approval.

It is preserving an independent ability to refuse execution.

# Execution-Boundary-Qwen38-27B-Q4_K_M

A Havenlon-focused Qwen3.8 27B model for reasoning about execution boundaries, Adversarial Completeness, independent veto, evidence chains, and the control of AI-driven real-world execution.

## Focus

- Execution Control
- Execution Boundaries
- Adversarial Completeness
- Independent veto
- Intent-to-Execution consistency
- Evidence chains
- Hardware-backed execution control
- AI Agent execution risk
- High-risk automated systems
- Fail-Secure behavior

## Run

```bash
ollama run Havenlon/Execution-Boundary-Qwen38-27B-Q4_K_M
```

## Quantization

Q4_K_M

## Core Principle

Havenlon assumes that software systems, AI Agents, APIs, accounts, policies, and approval workflows may eventually be misused, compromised, or operate on incorrect information.

Security therefore cannot depend only on preventing compromise.

It must also limit what a compromised component can ultimately cause.

A local failure should not automatically become a final real-world action.

The execution boundary exists between:

**what software wants to happen**

and

**what reality is allowed to become**

Its most important capability is not producing another approval.

It is preserving an independent ability to refuse execution.

Paste, drop or click to upload images (.png, .jpeg, .jpg, .svg, .gif)