AI Policy

• Development: AI-assisted analysis, boilerplate, refactoring, rapid POCs.
• Documentation: AI-generated readmes and auto-docs with manual review / edits.
• Articles: Blog posts and articles are my own words; AI is used for summaries and review.
• Communication: All human.

I own what I build.

The bot should not speak or decide for me.

If an agent is acting on my behalf, this will be declared.

I will not use the bot to speak in my voice.

A pen is still a valid thinking aid.

AI Tooling

I run local models with a full private tool chain, supporting Qwen 3.6 @ 35tok/s and 3.8 @ 45tok/s (all performance improvements negated by waaaay too much talking) with a context window of 128k, this handles day to day development tasks across multi-repo work.

I maintain my own Agent Sandbox with locked down permissions and my own development tool chain baked in running the pi.dev agent harness.

All data, plans and client information remain local - no third parties.

Why

I feel that in the current world, we can’t avoid AI, but we should also avoid completely abdicating ourselves to it. So in the rush to remove all friction for maximum speed, we should remember that friction still serves a purpose, and with speed comes increased accountability and responsibility.

In order to seek out and collaborate with other like minded people, we can only do this if we know where we sit ourselves. So this is my AI policy, and how I define my relationship with the bot, my thoughts, and what I bring to the table.