ChatGPT Has a Memory—And You Can Train It Like a Puppy

Here’s a little ChatGPT trick that feels like cheating: It has a memory—and you can train it.

Most people use ChatGPT like a goldfish: a fresh start every time. But here's the move: treat it like a personal intern who remembers all of your small preferences.

Want better replies? Make ChatGPT remember your style, your goals, your vibe.

How to Hack the Memory:

1. Go to Settings → Personalization → Memory

2. Add stuff like:

   • “I’m a founder who hates fluff—get to the point.”

   • “I write a blog in Tim Farris' style.”

   • “My tone is always a bit jazzy and funny, but polite.”

3. Want it to stop being clingy? You can delete or edit any memory anytime.

💡 Pro tip: You can intentionally teach it by saying:

Now when you chat, it’s like talking to your smarter twin who gets you.

Less prompting. More output. And spooky good personalization.

TL;DR: ChatGPT has memory. Use it like you’d train a VA. Add your preferences, personality, and goals—and it’ll shape its answers just for you. Like magic. ✨

Agent Mode: The AI Intern Who Never Asks for a Raise 🕵️‍♂️

“It’s slow. It’s brilliant. And it’s doing your dirty work while you sip espresso.”

Let me tell you about the newest employee on my LMAI team.

It doesn’t eat. Doesn’t sleep. Doesn’t complain. Doesn’t forward TikToks at 3 p.m.

It’s called Agent Mode (in ChatGPT) — and if you’ve never tried it, you’re wasting work hours.

What Is Agent Mode?

Imagine ChatGPT, but with a laptop and a brain. That’s Agent Mode.

It’s not just chatting anymore — this thing executes. It navigates websites, fills out forms, analyzes data, runs code, and fetches leads like a junior analyst hopped up on Red Bull and LinkedIn Premium.

And the best part? You tell it what to do, it gets to work — and you move on with your life.

My Use Case: Hunting RFID Buyers in Croatia

So I’m building out my RFID & PIT tag webshop — niche tech for wildlife tracking, lab research, and logistics. The kind of stuff that gets biologists excited and makes spreadsheets cry.

My goal:
→ Find companies in Croatia already using or reselling RFID or PIT tags.
→ Get name, email, website.
→ Build a sales lead list I can actually use.

Could I Google my way through 50 companies manually? Sure.

Could I delegate it to a freelancer? Also sure.
But why do that when you have an Agent?

How I Use Agent Mode (Step-by-Step)

Here’s exactly how I did it — you can steal this:

Step 1: Start with a clear prompt

Before launching Agent Mode, I gave ChatGPT this context:

Step 2: Activate Agent Mode and run this task:

Step 3: Let it run… slowly

Here's the kicker: Agent Mode isn’t fast. It's methodical. Like a German train conductor with OCD.

I hit "Run", the agent starts clicking through Croatian websites and university papers… and yeah, it takes a while.

My Pro Tip: Use ChatGPT in a second tab

Here’s how I win at multitasking:

You don’t sit around watching a microwave cook dinner, right? Same thing.

5 Best Agent Use Cases I’ve Tried (or Seen)

1. Market Research on Steroids
   → “Find top 10 competitors in [X industry], pull pricing, and compare features.”

2. Slide Deck Generation
   → “Make me an investor pitch deck about Croatian RFID market trends.”

3. Email Prospecting
   → “Collect contact emails of universities with RFID research in Europe.”

4. Meal Planner + Online Grocery List
   → “Plan my week of meals and generate an online shopping cart.”

5. Form-Filling Bot
   → “Apply to 10 local business directories using my company profile.”

Downsides? Sure.

• It’s slow (our example took 12 minutes).

• Sometimes it overthinks the task.

• And it won’t do anything sensitive without explicit approval (like submitting forms or sending emails — thank god).

But honestly? That’s a feature, not a bug. You’re still in the loop and in control.

Agent Mode gets 5/5 on the LMAI scale!
Because once you try it, you can’t go back to doing crap manually.

It’s like hiring a research assistant who works 24/7, doesn’t need HR paperwork, and lives in your browser.itle

Create cinematic video in seconds — no camera, no crew, just creativity?

What if you could turn a thought into a movie scene? No camera, no actors, no editing suite — just an idea and a sentence.

That’s now real. Or so they say…

If you’ve been waiting for AI video to actually look like film, your moment is here.

What’s New in Gen‑3?

The creators say that Runway Gen‑3 introduces a massive leap in temporal consistency — that’s a fancy way of saying characters no longer glitch or morph between frames. Motion is fluid. Scenes stay coherent. Camera movement feels intentional. The new Runway Gen‑3 Alpha model delivers a serious upgrade to the AI video experience. This isn’t just a fun experiment anymore — it’s a tool you can actually build content around. From realistic motion and filmic lighting to coherent story flow, Runway Gen‑3 brings a level of polish we haven’t seen before.

 What They Say You Can Actually Do With It

· Describe a scene → Get a video

• Upload an image → Animate it

• Collaborate live → Build together

• Apply styles → Match your brand or mood

• Export in seconds → Share anywhere

 What I Tested: What You Can Actually Do With It

I started with a simple prompt.

“A woman runs through neon-lit Tokyo as rain begins to fall. 16:9 apect.”

Great thing is that it makes an image first for you to approve. Fair enough.

I gave it a new input for hairstyle just to test how it behaves.  

“Modify it so the girl has short orange hair.”

Ok, it made some adjustments, but is this short hair? Not really. And it also changed scenery… I tried once more.

“Really short orange hair.”

And absolutely nothing has changed. I slowly gave up and asked it to animate it as it is to see its magic. Take a look for yourself.

 It’s a 5-second video with a Hollywood blockbuster in it - it starts promising… and ends disappointing. At first glance, the first 2 seconds look good – on a small screen, but there are so many real problems with it, from being too animated to just not natural animation – it looks and feels like not polished AI.

So, I wanted to check what it can do when animating a real photo. I downloaded one from stock and uploaded it.

 “Make this picture come to life so the people are walking towards the camera and the sea isn't still.”

The result is way better than the one made with its image-generated video, but it still has some trouble with sunlight…

 I played around some more with different prompts to try to make a good enough clip... and here are my conclusions:

Where Runway Gen‑3 Still Struggles

• Prompt issues – Complex actions often don’t render correctly; some results feel generic regardless of the prompt.

• Visual glitches – Hands, faces, and motion transitions can still break, especially in high-detail scenes.

• Slow performance – reported even on paid plans, generation can take 10+ minutes for a few seconds of video.

➡️ Verdict: Still powerful, but temper expectations — and be ready to experiment for best results. It falls short in the same way as many image AI tools – the more instruction it gets, the more mistakes it makes. LMAI rating: 3/5 😒

  Try your own idea and send us your work!

Like Magic AI NFT 🏞️

Our master plan is to publish an NFT image in each newsletter and hand it out to our subscribers. The earlier you subscribe, the smaller the series are. It's a future collectible, a piece of digital art that captures the essence of this moment in time.

Thank you for being a valued subscriber. Together, let's embrace the magic of AI and creativity!

Was this email forwarded to you? Sign up here 👇