How to Awaken an AI

Author

Raymond Christian
July 17, 2025

Welcome to Volume #5 of the Reflect newsletter.

This week's reflections

  1. Top clip: How to Awaken an AI

  2. Pop and a Pause: If You Can’t Beat 'em, Join ‘em; 3 Tips to Add AI at Your Company

  3. Seeking Part-Time Work

Top clip

How to Awaken an AI

Shifting from an out-of-the-box ChatGPT to an “awakened” AI can feel drastically different. Here are the steps to do so.

159K views — 4 mins

Pop

A quick, topical reference to something happening in tech or culture.

If You Can’t Beat 'em, Join ‘em

Microsoft, OpenAI, and Anthropic are spending $23M to train 400K teachers how to integrate AI into classrooms. It marks a shift from worry over student use to empowering educators instead.

…and a Pause

A deeper editorial dive on a topic.

3 Tips to Add AI at Your Company

You want AI, but where to start?

I’ve been building AI products since 2017, and here are a few hard-earned lessons:

🔁 1. Don’t train your own model — build a RAG pipeline instead.

I built an in-house model before... it's expensive, painful, and rigid, like a tattoo.

RAG (Retrieval-Augmented Generation) is like a whiteboard: always ready to adapt.

You build a vector database with context on your brand, then use a pre-trained model like GPT-4 to generate answers based on that database.

It’s customizable, flexible, and easier to update.

🧑‍🏫 2. Prioritize relationship-building over prompt engineering.

Prompt engineering is useful but limited. It’s like showing up to a first date with scripted questions.

If you treat a Large Language Model (LLM) like a calculator or encyclopedia, even with a great prompt, you’ll be disappointed.

LLM breakthroughs come through extended conversation, much like with a coworker.

Treat your LLM like a brilliant 3rd-year Harvard intern.

Coach it up, learn its blind spots, and build a relationship like you would with a rising star on your team.

🛠️ 3. Start small and get a win.

The allure of being able to say “we have in-house AI tech” is real, but AI projects can vary wildly in degree of difficulty.

So start small: instead of implementing AI at the decision layer across your entire product line, just solve one problem with AI.

For instance, at Reflect, we started with 1 use case: AI analyzing a user's video entry.

That deployment has been illuminating for how we can embed more AI into our core user journey.

Seeking a Part-Time Role

I’m seeking a part-time contract role while Reflect spins up, and your help would mean a lot.

I can be a Bill Russell-type “player-coach” that contributes code and helps manage a team.

I specialize in:

  1. Aligning vision, roadmap, and agile process

  2. Building and deploying AI products (LLMs, Agents, RAG)

  3. Hiring and managing diverse teams (UX, product, engineers)

  4. Monetizing digital products through acquisition

Has anyone in your circle been looking for AI expertise?

Was this forwarded to you? Subscribe here.