Steve

The problem

Career plans unfold over months, while most AI conversations begin without the decisions and context that came before. I wanted a thought partner that could keep the whole picture in view without turning every conversation into generic career advice.

The approach

I built Steve around a durable strategic plan, a current decision log, and explicit rules for how it should challenge me. Its job is to preserve context, make trade-offs visible, and help me pressure-test important choices while leaving the final judgment with me.

How it worked

Steve reads a compact set of files covering the long-term direction, the live plan, and the latest developments. It helps with weekly reviews, tracks dated commitments, pressure-tests consequential decisions, and supports drafts when I ask for them. It cannot contact anyone or act externally.

Key decisions

• Base the character on the spare, late-career Steve Jobs rather than the volatile caricature.
• Separate durable context, the live plan, and the recent log so old assumptions do not override new facts.
• Keep external actions outside the agent's authority and preserve my role as the principal making the decision.

Outcome

Steve remains in active use as a private planning and decision-support system. It gives longer-running career questions continuity and creates a written record of what changed, what was decided, and what still needs attention.

What I learned

A strategy agent becomes useful through memory design and behavioral constraints, not through a famous persona alone. The most important choices were what context it should retain, when it should challenge me, and when it should stop talking.