Hi all,
I had a call this week with a friend in the industry who asked me where I think AI is taking our business. This is what I told him.
Nobody has the crystal ball
I’m experimenting a lot with AI. But a lot. For the last 2 years I’ve been trying to get smart about everything I see.
I don’t have the crystal ball. And whoever says that he or she has it, and that in 5 years we’re going to be here, is probably lying. Because we don’t know what we’re going to do in 2 weeks.
Any video from the gurus out there that is older than 2 or 3 weeks is probably outdated already. The developments in all platforms are so impressive that you have different capabilities every day.
What is going to be automated
What I’m seeing is that there are certain tasks that are going to be automated with agents in the next 2, 3 years.
Editing is one of them. QC in the backend is another. And design, at least for our needs.
The WSC Sports of the world have been doing this for a while (cutting highlights automatically for sporting events and with very good outputs that millions of people consume on traditional tv formats or social platforms). They are getting better, but that is becoming a commodity.
Because in the past, you needed all this amount of money to develop an MVP (minimum viable product). Nowadays, if you’re smart, you can vibe code a basic product, present it to a couple of clients, and be up and running in a week. Then you can raise the money in a month. And then you could be out of business in a year, or a millionaire (or billionaire) in a couple of years.
To have a team at the beginning of any POC, you needed $3 million to start. Nowadays, if you want to burn tokens, you can burn tokens with a $100 Claude Max subscription.
Change your mindset
The traditional editor that tells you “I have been using Premiere for the last years and that’s the only thing”? If that guy doesn’t change his mindset, he will be out of work very soon.
Because any given person, who’s not an editor but understands a little bit of editing, who’s not a designer but understands a little bit of design, who’s not a filmmaker but understands a little bit of video production, now has superpowers.
It started with a post by Andrej Karpathy, who challenged us to ask AI for different kinds of outputs, not just text.
If you follow what Andrej proposes there, and you go and tell whatever model you’re using: “I need you to create for me a video explanation using ASD-STE 100 as a concept,” the quality of the video that thing is going to produce is 10 times better. And faster.
For instance, this is a video I created with Claude in less than 30 minutes. I analyze why an almost 8 minute clip about the housing problem in Spain went viral on multiple platforms.
In the past you needed to invest $30,000 to get one video like that. If you don’t know where it comes from, you would say this guy took a week editing it and paid a designer to do the design.
By the way, here is Jorge Galindo’s viral video. Take a look at it even if you don’t understand Spanish. It works even without audio, and that is the magic.
A team of agents
Let’s use a baseball example. Maybe not for MLB yet, but everything minor leagues.
We’re going to be able to grab every game and tell our team of AI agents: “I need this for every game. Look at me editing this first version. From now on, I want you to edit every game that hits this AWS bucket.”
As soon as it hits it, you will have a team. Your editor, your producer, your color corrector. An agent evaluating, QCing, the work of the others. And an agent that is just posting.
And then you’re going to get just the last message: “Hey, the 23 minute condensed version of the game is done.”
Record yourself
If you are not recording yourself while you do what you’re doing, you’re wasting time. You could replicate and superpower yourself by passing this information to your agentic team.
Start with one game. You record yourself, or you give the agent the possibility of seeing what you’re doing and hearing from you. Then, an AI agent is going to start creating a map. Okay, this is what good looks like.
Then the agent is going to say: “I’m going to cut the intro, because the intro is too long.” I just want “welcome to the game between the Miami Sharks and the Wisconsin Cheeses”. Boom, here we are.
Maybe we tell it: if there is a rendition of the national anthem, please show at least 5 seconds of the person singing it.
And then it’s going to start learning to include that always in the highlights.
Or maybe the first cut is done by an agent, and the editor enhances it from there.
Humans are still responsible
There is something that has been sticking in my head since I heard it from a guy last month.
The main difference where we are right now is that we, humans, are still responsible for everything that we put out there.
If I’m contracting you to condense games to 23 minutes and I’m paying you, the agent is not the responsible one. You are. So at some point there is a human who will greenlight the product, and review it at the end before it goes to the client.
You’re going to sit down, review it, and say: “Okay, this is good.” Or: “Change A, B, C and D.” But forget about the manual work for the rest.
Lower tiers first
I don’t think we are seeing this for a Super Bowl or an NBA Finals in the next 4 years, probably.
But we are already seeing this for everything that is lower tier.
This is it for this week.
But before leaving, if you did not understand some or all of the concepts that I talked about in this edition, click on this link to have them explained to you.
The Ministry of Broadcasting AI & Tech Glossary
This is not knowledge that only IT kids or engineers need to learn.