How to Get Better Results From AI

After 1.5 years running an AI astrology chat platform (and many customer support requests), here's something I've learned:

People who at least sort of understand how AI works (as a technology) get the best results. And people who get "bad" results...suddenly get good ones after changing a few simple things.

So today, I'm going to share two things:

1. A quick snapshot of how AI works
2. 3 quick tips on how to get the best results

Want to skip the tech lesson? Scroll down to 3 Ways to Get Better Results From AI

The Quick Answer: How Do You Get Better Results From AI?

If you want better results from AI tools like ChatGPT, Claude, Gemini, or Star Path, give the AI less irrelevant context and clearer instructions.

Here are the three best practices that help:

The basic principle behind all three?

More context isn't always better context.

Now let's talk about why.

How Does AI Work? AI Is a Pattern Matcher.

When you talk to an "AI", you interact with something called a "Large Language Model". In human-speak? It's basically a big engine designed to predict the next most likely word.

But you can change the way the engine behaves in a few important ways:

1. AI Models Are Trained on Data, Which Gives Them Patterns to Match

You can "train" AI on different data (like websites, articles, books, code, and other text), which basically gives the engine patterns to match.

That's part of why different AI models can sound (and behave) quite differently.

It's also why you get a different flavor from ChatGPT vs. Claude vs. Gemini. Different models are built and trained differently, so they speak—and pattern match—in different ways.

2. You Can Give AI Instructions That Change How It Responds

You can also give AI different instructions, which changes how it responds.

AI instructions can change things like:

• Tone
• Structure
• How detailed an answer should be
• Specific behaviors around a task, like analyzing a birth chart

One type of instruction is called a system prompt.

It's the thing you don't see behind the scenes at Star Path (or platforms like ChatGPT or Claude) that helps tell the AI how to behave.

Now, I've worked with a LOT of different AI models at this point in Star Path's evolution.

And let me tell you: their behavior can change massively when you add the right prompt.

But you also can't fundamentally change every characteristic of a specific model with prompting—because you're still working with the same underlying model you started with.

Here's a practical example:

Some earlier ChatGPT models became known for being extremely compassionate, empathic, and realllllly encouraging.

Too encouraging, actually.

In fact, that type of excessive agreement became a recognized AI safety problem because an AI that constantly validates its user can reinforce beliefs it shouldn't reinforce.

Newer AI systems have increasingly been designed to push back, point out blind spots, and avoid just...agreeing with everything a user says.

Now here's the cool part:

3. Your Messages Also Change How AI Responds

Ever noticed that if you talk to ChatGPT, its tone changes the longer you chat?

That's because the conversation itself becomes part of the context the AI is responding to.

If you use an emoji, the AI might start using emojis. If you write casually, it might start to sound more casual.

So in effect...

You become part of the pattern.

That's why Star Path works so well as a mirror. It "pattern matches" your birth chart and your current experience (which you share via messages).

So, simply by talking to AI, you change how it behaves.

But here's the catch:

If you aren't practicing good AI "hygiene," you can accidentally give the AI context that makes its answers worse instead of better.

3 Ways to Get Better Results From AI

Here are 3 "best practices" to help you get better results when chatting with AI.

They'll work for Star Path or other AI platforms like ChatGPT, Gemini, or Claude.

Tip #1: Avoid Super Long Messages and Unnecessary Chat History

Avoid super long messages when you don't actually need all of that information, especially things like pasting full chat histories.

Instead, opt for quick summaries.

Long messages = harder pattern matching. The more an AI has to comb through, the more likely it'll "miss" something (or give you an output that just doesn't hit the mark).

Keeping it short optimizes your energy use and gets you better results because the longer the message gets, the more "junk memory" you unintentionally bloat your chat with.

Tip #2: Use a Fresh Chat When You Need to Give AI a Long Document

Sometimes, a long input really is necessary.

Maybe you're pasting:

• A podcast transcript
• Notes from a client meeting
• A long article
• A document you want analyzed

If you need to paste something long, start a fresh chat and be very specific about the task you need the AI to perform.

The longer the input, the simpler the instruction needs to be, otherwise AIs can get confused about what to prioritize via pattern matching.

This is especially problematic if you're pasting something long into an already-long chat.

Long document + long chat history = a lot of potential inputs competing for attention.

So if you need to send a long message, try using a fresh chat.

Tip #3: Curate AI Memory Instead of Keeping One Giant Chat

If you're working on a particular subject that requires memory, create a Digital Altar (i.e. a folder with cross-chat memory), then start fresh chats inside it.

Turns out, one of the biggest drivers of both energy use + "the AI is acting weird" complaints is...

Too much context.

And the truth is: your AI doesn't actually need to remember most of what you've discussed.

Most words in a chat are "filler." Short summaries can actually give AI a cleaner foundation for pattern matching than long, organically-phrased conversations.

In Star Path, we use Digital Altars to curate memory around specific projects. This gives you the benefits of simple (but targeted) memory + fresh chats.

Watch this tutorial on how to create a folder with cross-chat memory Or, read this article

Using another AI chat platform? You might be able to edit your memory, create projects, or add custom instructions. This varies by provider.

Click here to try these tips with Star Path

Fresh Chat vs. Long Chat vs. Curated Memory: Which Should You Use?

The Bottom Line: Better AI Results Come From Better Context

If there's one thing I've learned troubleshooting support requests based on thousands of conversations happening inside an AI platform, it's this:

More context isn't necessarily better. Better context is better.

AI is reallllly good at working with patterns. So your job isn't to give it everything you've ever said.

Your job is to give it the right information for the task you're asking it to perform.

So if your AI starts acting weird, missing obvious things, or giving you answers that don't quite hit the mark, try these three things:

1. Shorten your message to the info that actually matters.
2. Start a fresh chat when you're introducing a big new task or document.
3. Use curated memory for information you truly need referenced across chats.

Sometimes, getting better results from AI isn't about writing a longer, more complicated prompt. In fact, the newer AI models are smart enough that longer prompts get worse results (because they limit the AI's innate thinking power).

It's about giving the AI less noise.

Click here to try these tips with Star Path