9 Ways the AI Power Users Are Actually Working — And What You Can Take From It

If you're keeping tabs on how AI is actually evolving in practice — not the hype, but the day-to-day — Nathaniel Whittemore's The AI Daily Brief is worth your time. NLW (as he's known) puts out sharp, well-researched episodes that cut through the noise without dumbing things down too much.

A recent episode stood out: "9 Codex Tips From the Codex Team" — where NLW breaks down strategies shared by members of OpenAI's own team on how they use Codex, their AI development environment, in real work. The source material comes from developer and AI researcher Jason Liu, who documented these practices in detail.

Now — a lot of this episode is aimed at developers and technical builders. But what's striking is how many of the underlying ideas apply to anyone trying to get serious value from AI in their business. Here's our take: what stood out, what's genuinely useful, and what's mostly for the technical crowd.

---

---

What The Episode Is Actually About

The episode focuses on nine practices that Jason Liu — an AI researcher and developer — documented from working closely with OpenAI's Codex. NLW walks through each one and adds his own framing.

A fair chunk of it is technical. Things like browser access, pull request workflows, and coding environments matter more to developers than to marketers or business owners. But the thinking behind each tip? That translates broadly.

Here's what stands out — and why it matters beyond the developer world.

---

What's Worth Your Attention

Persistent memory is the biggest unlock most people are ignoring.

The tip about building "durable threads" and external memory files — basically, giving your AI a notebook that carries context forward — felt like the most underrated idea in the whole episode. Everyone complains that AI "forgets" things. The solution isn't a better model. It's building a system where the important context never has to be re-explained. Your brand voice, your customer personas, your business goals — document them once in a format your AI can access, and every output gets better immediately.

That one's practical for any business, starting today.

The "steering while it works" mindset is a real shift.

Most people prompt an AI, wait for the response, evaluate it, then prompt again. What the Codex team describes is more like managing a live process — you're giving direction continuously, not in a back-and-forth loop. NLW frames it well: it's less about prompting and more about operating. The faster that shift gets internalized, the more leverage you get.

Heartbeats and scheduled automations are where it gets genuinely powerful — and accessible.

Setting up recurring AI tasks that run on their own — monitoring inboxes, checking data, flagging changes — isn't as technically demanding as it sounds. And for business owners, it's probably the closest thing to having an extra set of eyes on your operations without hiring anyone. Of all the tips in the episode, this one translates most directly to a non-developer context.

The "set real goals with real ways to verify them" tip is deceptively simple.

At first it sounds obvious. But think about how most people actually use AI: "improve this email," "make this better," "rewrite this." Vague. The Codex team's approach is to define what done looks like before you start — and have a way to confirm it. That discipline improves outcomes immediately, and it has nothing to do with what AI tool you're using.

---

What's More Developer-Specific (But Still Worth Knowing About)

The tips around remote control, browser access, and computer use are genuinely impressive — but they require technical setup most business users won't do themselves. The concept still matters: AI can operate tools on your behalf, not just generate text. That's the direction everything is moving. You don't need to build it yourself today, but understanding that it exists changes how you should be thinking about where AI fits in your workflows.

---

The Honest Takeaway

The episode is good. NLW keeps things grounded and doesn't oversell. The source material from Jason Liu is thorough, and the translation into a 29-minute listen is well done.

The main takeaway: the gap between people using AI casually and people using it as actual infrastructure is growing fast. These nine strategies aren't magic. They're just what it looks like when someone builds a real system instead of using AI like a one-off tool.

The businesses that figure that out first are going to have a meaningful advantage. Not because they have access to anything others don't — everything discussed in this episode is available to anyone. It's just about whether you build the habits and systems to use it seriously.

Worth the 29 minutes.

---

All credit to Nathaniel Whittemore (NLW) and The AI Daily Brief for the original episode. Source material drawn from Jason Liu's Codex Maxxing guide. This post reflects our own interpretation and perspective — go listen to the full episode for the complete picture.