11 Comments
User's avatar
Daria Cupareanu's avatar

Fantastic guide, Tam! I completely agree, even when you provide GPT with some context, it often falls short of capturing the depth and specificity your work requires. Adding detailed information about your business, goals, and clients makes a huge difference. It also helps eliminate the constant stress of figuring out what to write next, since we often overlook the expertise we already have while searching for something new or novel.

Expand full comment
Tam Nguyen's avatar

Exactly it’s chaotic creation vs systematic extraction. We have have valuable conversations in passing all day. It’s just about extracting ideas.

This goes for newsletters as well - which I’ll be looking forward to building on soon.

Tough part is automating with tools like make or n8n because you can give it that much depth. Too costly.

Expand full comment
Renzo Alvau's avatar

Interesting concept, Tam. However, the challenge lies in training the AI to understand nuanced language and context. Also, how does it handle the vastness of data while ensuring low latency? Would love to hear more about your model's architecture and its scalability aspects.

Expand full comment
Tam Nguyen's avatar

Totally fair questions and I should clarify: I’m not training a model or building a production SaaS tool here. This is more like setting up a smart assistant (Claude or GPT) that knows your business.

It works one transcript or voice note at a time, guided by a “Business Foundation” doc (offers, ICP, goals, etc.) that keeps things focused.

That said, I’m definitely thinking about how to make this more efficient for solopreneurs and small teams - especially around the context setup and automating the idea extraction process. For now, relevance > speed.

Expand full comment
George Siosi Samuels's avatar

trying out now

Expand full comment
Tam Nguyen's avatar

How’d it go?

Expand full comment
George Siosi Samuels's avatar

Nice doc towards the end. But needed more specifics on how to apply to newsletters or LI posts via instructions on a GPT folder etc. didn’t translate as smoothly when taken out of the convo thread

Expand full comment
Tam Nguyen's avatar

That’s great feedback - thanks George.

You may be interested in the final post of the week. It takes into account multiple context docs. https://www.thecontentkitchen.co/p/the-200k-consultant-living-in-your

For newsletter writing and Linkedin posts you'll likely need a separate doc for each explaining how to write for each platform, templates, etc.

@dariacupareanu just wrote a great article on helping AI understand what 'good' looks like and how to create 'knowledge packets' for it. https://substack.com/@dariacupareanu/p-169439466

Expand full comment
Mike Goitein's avatar

Powerful stuff, Tam

And that illustration is spot-on

The notion of a “Business Foundation document” is so key to provide the unique business context necessary to create content that converts.

Kudos!

Expand full comment
Tam Nguyen's avatar

Thanks Mike. Are there any key pieces of info that you swear by to train AI?

Expand full comment
Mike Goitein's avatar

Sure, Tam!

I've had the best success with using my most highest-performing articles & newsletters.

LLMs always need context & examples!

Once they get that, they're able to provide WAY better answers...

And the other thing is I rely on the responses more as editing feedback than as final copy.

It's crucial for me to rewrite everything, both as part of my writer's craft to continue improving, as well as for making the final product I put out there truly mine.

Expand full comment