Our experience using Qwen3.8-27B, Pi and LM Studio on an ASUS Ascent GX10 for real development, followed by cross-review with Codex and production validation.
AILocal ModelsCoding AgentsOpen Weights
Read →Insights
Notes from work on real systems.
Practical guides, case studies and reflections from the XSEVEN team on AI, software engineering, DevOps and infrastructure.
11 articles
Qwen3.8-27B and Pi: A Local LLM Agent at Work
Our experience using Qwen3.8-27B, Pi and LM Studio on an ASUS Ascent GX10 for real development, followed by cross-review with Codex and production validation.
ReadVibe Coding and AI: Working Software Is Not Always Governable Software
A reflection on software in the age of AI: building working products is becoming more accessible, but production, security, maintainability and responsibility remain real engineering problems.
ReadLocal AI and Uncensored Models: Technical Freedom or Risk Without Friction?
A reflection on local uncensored AI models, from hands-on experimentation with personal AI hardware to LM Studio, llama.cpp, safety, responsibility and unfiltered access to potentially harmful capabilities.
ReadClaude Fable, Codex and the Duty to Question AI Rankings
Starting from Salvatore Sanfilippo's observations on Claude Fable 5 and GPT 5.5, a critical reflection on benchmarks, coding agents, project context and AI model evaluation.
ReadThe Invisible Work of Sysadmins: Why Infrastructure Is Not a Playground
A reflection on the often underestimated role of Sysadmins, DevOps and SREs: responsibility, troubleshooting, scale, automation, AI and necessary boundaries in shared infrastructure.
ReadThe Psychological Impact of AI: Learning Not to Humanize Models
Generative AI can sound convincing and empathetic, but it does not feel, intend or care. A reflection on LLMs, anthropomorphism, education and responsibility for generations growing up with AI.
ReadWhen AI Builds Itself: Governance, Power and Competitive Risk
After Anthropic's article on recursive self-improvement, a reflection on AI governance, credible pauses, company competition, government power, risks and possibilities.
ReadPrivate AI: Turning Enterprise Knowledge into Governed Intelligence
Private AI is not just running an LLM on private infrastructure. It is the discipline of connecting AI to enterprise knowledge, processes and governance while keeping control over data, access and compliance.
ReadCloud Repatriation: Why We Moved a Client from AWS to Bare Metal
How we migrated a production workload from AWS to Leaseweb bare-metal servers, cutting costs from tens of thousands to a few thousand euros per month while improving reliability, performance and operational control.
ReadThe DevOps Philosophy: Why Your Company Should Adopt It
From throwing applications over the wall to shared ownership of the entire delivery lifecycle — what DevOps actually means, why it works, and what changes when an organisation genuinely adopts it.
ReadKubernetes Cost Optimization: a Practical Guide for 2025
A practical Kubernetes cost optimization guide: namespace cost allocation, right-sizing, VPA, HPA, spot nodes, Karpenter, quotas, cleanup and FinOps routines.
Read