🔥 AI's 10M Token Leap: The Marketing Edge Nobody's Talking About

This week, a colleague complained about ChatGPT's "failure" to consistently (85%+) complete dozens of complex analysis tasks.

🧐 My first thought: Would we hold junior human analysts to that same standard?

This coincided with Stanford releasing their 2025 AI Index Report, which contains eye-opening data on AI's progression. The cost reduction to use AI between 2022-2025 is staggering, while performance has improved dramatically across everything from PhD-level science to language understanding.

The drop in our cost-to-use AI is staggering… 

At the same time, AI’s gotten significantly better at everything from PhD-level science to understanding the English language... ☕️

And image creation is no exception to the trend of relentless improvement...  

BUT, we still have a loooong way to go before it takes over completely from us humans. 👀 

And we have lots of work to continue to do on safety and responsibility. 

More on safety and responsible implementation of AI in the workplace, in our guide to AI compliance checklist.

📌[A HANDY RESOURCE] The AI Compliance Checklist: A Law-Abiding Marketer’s Guide to Using AI the Right Way (The First Time)

🫠 🧠 What really melted my mind this week, was Meta Llama's release of a 10M token context window.

I've been talking about an infinite context window for a bit, but to see it arrive so soon was shocking…

…As a reminder, GPT-4 was released in June of 2023 with an 8,192 token window. 

The Bottom Line: What All This Actually Means For Marketers 🧠

Let’s think about the implications.

Today: Upload 20 PDFs. Tomorrow: Upload 1,500-1,600 PDFs.

This means loading:

• All competitor ads

• All your ads

• All competitor landing pages

• All your landing pages

• All ad copy (yours and theirs)

Producing one hell of a competitive ad analysis.🫠 

Let's assume (conservatively) that this doubles every 8 months.

We are talking 196,923 PDFs of 5,000 words each into a 1.28 billion-token context window in a few years.

👉🏼 Let's assume we track 40 competitors a month. We could load 410 years of competitive information into that context window (you would need to be one of these 5,568 companies for that to be useful). 

Most marketers are not thinking about how to use AI with an effectively unlimited context window. 

I want you to. 

Your marketing machine will be starved for data to feed the AI. 

Where are you going to get it? 

How are you going to make sure it's accurate? 

Who's going to QA it? 

You need to be thinking about data pipelines at scale. 

Maybe that's SEO data. Maybe that's ad data.

It could be all of it. 

Because let's be honest – nobody wants to manually manage thousands of PDFs monthly.

Do you?

⚡ Quick Win

✅ Start a list of your most valuable marketing data sources today. What competitive intelligence would transform your strategy if you could analyze it all at once?

✅ Test using ChatGPT unlimited chat Memory to Plus and Pro Accounts that they added Yesterday.

Happy Friday! 

P.S. See you at Ask Alec next week! Where we go deep on tools.