AI Power-Users Leverage Context More Than Skill
Most people open a new AI chat, get decent results, and wonder why the output never quite sounds like them. Axios co-founder Jim VandeHei's answer is simple: you haven't taught it who you are. In a two-part Finish Line series, he offers a practical framework for getting significantly more out of AI tools, starting with prompting and building through memory management and project workspaces.
On prompting, VandeHei covers four moves:
- Give the AI full context upfront including your role, audience, and intent
- Ask it what questions it needs answered before it starts
- Show it examples of work you want to replicate
- Push back on the first draft rather than accepting it
His suggestion for building reusable prompts is worth noting: ask the AI to reverse-engineer the output you liked into a template you can paste at the top of any future request.
On memory, the distinction between working memory and lasting memory is the practical core:
- Working memory resets when you close a chat
- Lasting memory, supported on paid plans, stores your preferences, role, and style across sessions
- End strong chats by explicitly telling the AI what to save
- Audit stored memories every few weeks
- Use past chat search to identify patterns in how you think and where you get stuck
For recurring work, VandeHei recommends project workspaces over standalone chats. Drop in past work, define your audience and tone in plain language, and the tool gets sharper the more you use it.
Full story: Axios Part 1 - Part 2