Skip to content
Week of May 19th
Try to scrape all of the videos from an ads library and then put that into Poppy to see what the holes are.
marketing thoughts filmed from far away
Type of CTAs: Specific Guide, Broad Guide, Link to Tool
Post 1: Make AI-Enhanced UGC Ads
Short Form Script:
Post 2: Make Your Own Useable Font
Make your own usable font with Chat GPT
Google Just Built an AI That’s Smarter Than We Thought Possible. (I think Google will outperform any other competitors by July in all fronts)
Here’s some AI Advertising News
Google deepmind just released Alpha Evolve on May 14th, an AI system with agents that don’t just solve difficult tasks and problems, but also improves itself nonstop. It writes codes, keeps the winners, tosses the losers and builds new versions from the best performers.
And this includes it’s own training. It’s found better designs for AI computer chips, helped google make their data centers more efficient saving them millions, and it’s beating out complex math techniques that haven’t been improved in 50 years.
It might be the first step towards AI that can actually think.
What does this mean for paid ads?
Well, modern ad platforms are powered by incredibly complex algorithms – for bidding, for targeting, for deciding which ad you see when. AlphaEvolve is the type of system that could make THOSE algorithms exponentially better.
And for ad creative, we’ll go more in the direction that Advantage+ and Google’s AI Max is taking us with more generative creative, automated testing, less human inputs.
My prediction is that the game of improving your performance will shift towards controlling the quality of the inputs you provide the AI and the strategic choices you make here.
Strategists and creatives will define the "why" and the "what," and AI will handles the "how" and the "at scale" execution, which means people in that more execution-focus role need to start developing skills in strategy.
I also think we’ll always need to manually monitor and course correct decisions that AI makes, so something like “AI Managers” might be a position that will pop up eventually.
These are just my predictions and I’m curious what you all think, and let me know if you have predictions of your own. I’m not saying I love this direction... but I really believe that we have to try and get ahead of it all and start developing future proofing skills.
Human curation and refinement of AI-generated creative will also be key to maintaining brand authenticity
Follow me to get more AI ad news like this! See ya.
Differentiation will increasingly come from the quality of the input you provide the AI and the strategic choices you make about how to use AI. This includes:
The most effective campaigns will likely be a collaboration between human strategists/creatives who define the "why" and the "what," and AI that handles the "how" and the "at scale" execution. Human curation and refinement of AI-generated creative will also be key to maintaining brand authenticity
Leveraging unique first-party data to inform personalization and creative generation.
Defining a distinct brand voice and personality that the AI is trained to adhere to.
Developing innovative campaign concepts and narratives for the AI to execute variations on.
Choosing which metrics to optimize for and how to weigh different objectives.
Imagine feeding it high-level goals and just a few starting creative elements... or maybe eventually, just the goals themselves. This AI could potentially evolve the actual ad creatives – constantly generating variations of headlines, text, even visuals – automatically testing them in live campaigns, learning from performance data, and continuously improving the creative shown to users.
You could essentially set the objective for creative performance, and the algorithm works to evolve the best possible ads over time.
This technology has the potential to build the next generation of ad tech, leading to campaigns that are not just optimized, but fundamentally more intelligent and efficient across targeting, bidding, and creative.
It means the tools performance advertisers use could become far more powerful, pushing the boundaries of what's possible in digital advertising.
Keep an eye on this – the future of ad optimization might just be evolving itself.
This process runs in a loop. It never stops improving. And it’s not just theory. Google is already using Alpha Evolve. It’s helped them make their data centers more efficient, saving them millions. It’s found better designs for their AI computer chips. It’s even discovered faster ways to do complex math, beating techniques that haven’t been improved in over 50 years. In one case, it solved a math problem in fewer steps than anyone thought was possible, and it did it in a way no human ever had.
The system works because it can test its own answers quickly and clearly. That’s why it’s especially powerful in areas like math, coding, and hardware design. It’s not ready for areas like biology or medicine where results can’t be tested instantly, but in the right settings, it’s a game changer.
t might be the first real step toward AI that can actually think.”
[Middle]
“This system doesn’t just follow instructions. It evolves. You give it a goal, like making code faster… and it improves itself nonstop.”
“It writes code, tests it, keeps the winners, tosses the losers, and builds new versions from the best ideas. Like evolution—but for intelligence.”
“It’s already discovered faster ways to do math, redesigned Google’s AI chips, and made data centers more efficient. Stuff humans haven’t improved in over 50 years.”
[Closing + Advertising Tie-In]
“Now imagine pointing that at advertising. AI that writes, tests, and evolves creative automatically—optimizing ad concepts the way nature refines DNA. That’s not far off.”
Google DeepMind has just introduced something that could change everything. An AI system called Alpha Evolve. What makes it different is that it can teach itself how to solve difficult problems and keep getting better without constant help from humans. It doesn’t just learn once and stop. It keeps evolving. It improves itself, thinks of new ideas, tests them, and keeps the best ones. And it’s already making discoveries that people couldn’t figure out for decades.
Here’s how it works in plain language. A person gives the AI a goal, like making a piece of code faster or solving a tricky math problem. Then the AI is given a way to measure how good its answer is. For example, if the goal is speed, the AI will test how fast the code runs. It also starts with a small piece of basic code, just enough to get it going. From there, the AI makes changes to that code over and over again, keeping the good versions and throwing away the bad ones. Each time, it builds new ideas from the best ones, like nature does in evolution.
This process runs in a loop. It never stops improving. And it’s not just theory. Google is already using Alpha Evolve. It’s helped them make their data centers more efficient, saving them millions. It’s found better designs for their AI computer chips. It’s even discovered faster ways to do complex math, beating techniques that haven’t been improved in over 50 years. In one case, it solved a math problem in fewer steps than anyone thought was possible, and it did it in a way no human ever had.
The system works because it can test its own answers quickly and clearly. That’s why it’s especially powerful in areas like math, coding, and hardware design. It’s not ready for areas like biology or medicine where results can’t be tested instantly, but in the right settings, it’s a game changer.
What’s also surprising is that Alpha Evolve doesn’t need brand new AI models to work. It uses the same tools available to the public, like Gemini 2.5, but inside a framework that keeps improving itself through constant feedback and testing.
Google has been quietly using this system for over a year. It’s already helped them improve internal systems, save energy, and speed up AI training. And if this is just the beginning, the future could get a lot more interesting. This might be the first real step toward AI that can discover things on its own. Not just follow instructions, but truly think.
Want to print your doc?
This is not the way.
This is not the way.
Try clicking the ⋯ next to your doc name or using a keyboard shortcut (
CtrlP
) instead.