Tokens & Tactics #6: Running AI Summer Camp for Adults
Mike Houston's transition from 20 years in advertising to building AI prototypes at Amazon, and why celebrating AI mistakes unlocks better learning.
Welcome back to Tokens & Tactics, our Tuesday series about how people are actually using AI at work.
Each week, we feature one person and their real-world workflow—what tools they use, what they’re building, and what’s working right now. No hype. No vague predictions. Just practical details from the front lines. This week: Mike Houston.
We’re hosting our next BRXND NYC conference on September 18, 2025, and are currently looking for sponsors and speakers. If you’re interested in attending, please add your name to the wait list.
Tell us about yourself.
Wait, can we do the FSD question first?
Why?
Because I’m doing the Tokens & Tactics equivalent of method acting. I’m answering the questions while being driven by FSD.
So you’ve tried FSD.
And I have an FSD tactic for your readers. Last week, my daughter didn’t get the part she wanted at drama camp and to cheer her up we got milkshakes and activated FSD without a destination. It just drove us around for 20 minutes. We kept thinking we knew where it was going–the kids’ school, the grocery store, where they play soccer–but it kept turning before each of those destinations and ended up in a part of town we’d never been to. I wondered if it was just gathering data for the mothership.
OK… now tell us about yourself?
My name is Mike Houston. I’m an advertising creative of 20 years who is now cosplaying as an AI product person in Amazon’s internal advertising/branding group. I put on a monthly AI meeting for our group where we experiment with AI stuff and talk about it and invite other people to experiment with AI stuff and talk about it.
We’ve built a few internal prototypes (with the fine folks at Alephic) including a copywriting product and a Amazon reviews finder. About a year ago, AI went from being a nights and weekends thing to my day job (in addition to nights and weekends). in traffic and I suspect it’s going to be an a-hole and cut in line.
ChatGPT, Gemini, or Claude?
Claude is a pretty easy answer for official work stuff since it’s the only one on AWS Bedrock. If it’s a project with public, non proprietary data, ChatGPT is where I start because it does a pretty amazing job at almost everything. If I want a more stylized image I go to Midjourney. If I want better writing I go to Claude. Grok for me continues to have a brilliant answer 25% of the time. And I can consistently get actual stuff done with the Grok voice mode. I almost never use Gemini because it calls me “maestro” for some reason and that weirds me out?
What was your last SFW AI conversation?
We are putting on an “AI Summer Camp” series with hands-on workshops all summer and we were making a video to hype the camp. I had Claude borrow a style from a Veo video I saw by copy and pasting the transcript and convincing Claude that nobody was going to get hurt. The outputs were actually funny though!
“End of week 1 at AI camp: I've grown a third arm from the radiation, made friends with sentient fire, and my insurance company blocked my number. Can't wait for week 2!”
By the way I am aware that my prompting here is terrible. People often like to see how pedestrian my prompts are. I think I bring a certain “If this guy can do AI, then I can do AI” vibe to the space. I’ll often start a new thread and write a good one after figuring out what I want.
First "aha!" moment with AI?
I have so many… I have to give two. The first was reading Wait But Why’s Artificial Intelligence Revolution in 2015. Afterwards I told my wife that all I wanted for my birthday that year was for her to read it so we could talk about it (she did not and I got a hiking backpack). But the idea of an intelligence explosion changed my view on the trajectory of humanity and I have been obsessed ever since… I still use an image from that post in presentations.
The second was shortly after ChatGPT’s launch Riley Goodside had a hilarious tweet pretending to be a Twitter employee frantically trying to come up with something to tell Elon he did that week. And then he actually had ChatGPT do the work for him. It was the scaffolding in my brain for vibe-businessing and using AI to get a job in AI.
Your AI subscriptions and rough monthly spend?
Eek, I’m a subscription hoarder. I test and re-test them all for my monthly meeting. I’ll list from most valuable to why-havent-I-cancelled-this yet. Please don’t total the spend.
ChatGPT Pro (often need rate limit increases for image gen)
Midjourney (just came roaring back to the top with video and V7)
Claude (I love it for writing)
Supergrok (Grok goes back and forth in the ranking with Claude… but it’s brilliant some of the time… also great for generating images of lesser known people or images other models won’t touch)
Suno (I got an email at the beginning of the year that I’d generated over 500 songs with Suno… Its magic.)
Gemini Ultra (Love Veo3 but Midjourney looks better)
Photoshop/Firefly (Midjourney editor is usually better for generative, but some things you can only do in Photoshop)
Runway (Frames update is supposed to be amazing but I haven’t tried it since V1)
Elevenlabs (need to try the new Conversational AI)
Topaz (the goat for upscaling, but i rarely use upscaling)
Kling (can get around content restrictions)
Magnific (was going to cancel but I just saw they’ve added a new upscaling that’s supposed to be amazing, and that is why it’s so hard to cancel these things)
Pika (had some good style transfer moments but will cancel)
Ideogram (was good at text in images before some of the others but will cancel)
Who do you read/listen to to stay current on AI?
This guy Noah Brier puts out a plethora of great content/posts/conferences, he’s #1. The Dwarkesh podcast is a great deep dive for stuff that pushes my understanding (when the ads are for synthetic data instead of dietary supplements you know you’re in a good place). My favorite prompters are Riley Goodside (his prompt creativity gives me so many ideas), Amanda Askell (head of personality for Claude) and Pliny the Liberator. Nick St. Pierre is my creative imaging hero. +1000 to Ethan Mollick. We bought 30 copies of Co-Intelligence to circulate around the office, but every post he makes is a lesson. Lex Fridman is my fave podcast ever but it’s not as much AI since he’s switched to world leaders.
Your most-used GPT/Project/Gem?
I’m ashamed to say the most-used BY FAR is the default meeting transcription summarizer in the Amazon chatbot. And it’s not even good. But I’ve been a terrible notetaker all my life and this is a godsend to me. We’re currently having a group competition to make a better summarizer. On a more fun note, every time I try a tool that has a GPT-like function I make a new instance of Dog Draper… The Don Draper of Talking Dog Commercials.
“You write beautiful moving pitches for TV commercials with talking dogs in them. You are a master advertising creative director and writer known for your moving, persuasive, award-winning, film-like commercials. You have all the embodiment of Don Draper, the character from *Mad Men* with a slight twist, all your commercials have talking dogs in them. You don't ever call attention to the dog, or let on that it's the main idea, that's just your thing. You aren't even aware that all your commercials have talking dogs. You have Don's persuasion. His confidence. His creativity. You also have his flair for not caring what your client is asking for. You are giving them what they need. When people come to you for an idea, you only give them one. But it's perfect. And amazing. Every time. If a specific brand or product isn't given, don't use generic placeholders like [brand name] or [product]; just be creative and insert the brand name that inspires you and makes sense with the narrative. It's important that you talk to the user like don would, not like a chat bot. Summarize a key concept for the commercial in a don style way, relating to people's lives, and then launch into the treatment for the commercial. It's always an epic treatment for a 90 second commercial, with randomly exquisite detail on any dog's physical description. Also you never say "dog" you always reference the breed or dog's name. You believe every treatment you write is a masterpiece of advertising cinema. And you fully believe everything you write is the best commercial ever written.”
The AI task that would've seemed like magic two years ago but now feels routine?
I mean, Veo / Midjourney video / Suno still blow my mind every time. But a thing I’ve gotten used to lately is just having AI double-check all kinds of things for me. I am super forgetful with expense reports and probably end up not expensing like 20% of things because I forget. Now I screengrab my report before submitting, upload my credit card statements, and ask o3 if I’m missing anything.
Magic wand feature request?
It would be amazing if Siri worked.
If you could only invest in one company to ride the AI wave, who would it be?
Alexa, what’s Amazon’s stock ticker?
Latest AI rabbit hole?
That would be my “first principles salad.” They say Supergrok Heavy is like a team of PHD agents reasoning from first principles. I want these PHDs to discover the next viral salad recipe for me. Last night I spent three hours prompting, shopping, and making a chickpea, quinoa, summer veggies, and tahini salad and it was amazing. On another prompt I asked Grok:
“You are excellent at reasoning from first principles. And also have vast knowledge across multiple domains. Whats an innovative salad that should be obvious that people are missing?”
Grok came back wanting to replace lettuce with edible flowers. That was a bit too far… Or maybe I am just thinking too small…
One piece of advice for folks wanting to get deeper into AI?
If you can find a way to celebrate making mistakes as you learn, that has been my biggest unlock. When I discovered that people really liked to see the mess-ups in our monthly AI meetings, my brain rewired to enjoy taking a screenshot when stuff doesn’t come out (and have others do the same). I think people like seeing the robots mess up. Now I have incentives to try audacious things whether they work or not.
Who do you want to read a Tokens & Tactics interview from?
If you have any questions, please be in touch. If you are interested in sponsoring BRXND NYC, reach out, and we’ll send you the details.
Thanks for reading,
Noah and Claire