Here's what's coming up:
ChatGPT is selling ads now (at $60 CPM, no less)
Meta wants to automate your entire creative team by Christmas
A man in Sweden is selling digital ketamine for ChatGPT
Grab your coffee. We're going in.
AD TECH
OpenAI flipped the ad switch about a month ago. Free and Go tier users in the US are now seeing sponsored products alongside their answers. (Nobody asked for this, but here we are.)
Criteo signed on as the first programmatic ad partner, bringing roughly 17,000 advertisers to the party.
The cover charge: $200K minimum commitment. CPM sits around $60. So no, you're not running a test campaign this weekend.
Ads and responses run on separate systems. OpenAI says ads don't influence what ChatGPT tells you. Make of that what you will.
Early (and not yet statistically conclusive) data suggests LLM-referred traffic converts at higher rates than many traditional referral sources. Which tracks, if you think about it.
Why It Matters
Here's what's different about this channel: the people seeing these ads are mid-thought. They typed a question. They're actively solving a problem. That's not someone doomscrolling Instagram at 11 PM in sweatpants.
That's someone with intent, attention, and often with a credit card within arm's reach.
That said, you won't be buying ChatGPT ads anytime soon (unless you've got $200K in couch cushions). But that's not the move anyway. The move is understanding that search has changed AI platforms are becoming discovery channels.
AI TOOLS
Meta's 2026 roadmap: full ad creation automation by year-end. Submit a product image and a budget. AI generates the creative, the copy, the video, chooses the placement, and optimizes the spend. You go take a nap.
New image-to-video tool converts up to 20 product photos into polished multi-scene video ads. Twenty still images in, video ads out. Just like that.
Meta's video generation tools already hit a $10B revenue run-rate in Q4 2025, growing nearly 3x faster than their overall ad revenue. They're betting the farm on this.
Meta Lattice (next-gen ML architecture) predicts customer behavior with 4x the accuracy of previous models. Manual audience targeting is getting automated, piece by piece.
Why It Matters
For a solo operator who's been stitching together Canva graphics and boosting posts with crossed fingers, this is either the best news I’ve heard all year or mildly terrifying. Both reactions seem correct.
The catch (because there's always a catch): automation runs on inputs. Good product photography. Clean catalog data. A value proposition the AI can actually work with. Feed it blurry phone photos and a vague description? You'll get blurry, vague ads. Automatically. At scale. Which is arguably worse than doing nothing.
Get your product images sharp. Get your data clean. Run a simple test campaign before the full suite drops, so you're not Googling "how does Meta AI ads work" while your competitors are scaling theirs.
⚡ SEE IT IN ACTION
Every story in today's issue has the same punchline: AI is handling the execution now. The only question is whether it's executing your strategy or someone else's.
The Claude Cowork Quick Start for Marketers. One video. Set up your AI agent the right way the first time.
Stop reading about the future. Start running it.
🧠 I'll take "AI Startups" for $800, Alex.
This startup emerged from stealth in February 2026 with $15 million in seed funding to build flexible AI agents that handle audience targeting, campaign management, and media planning for marketers.
Answer... ermm... question at the bottom of this email.👇
🤪 OH, AND THIS…
A Swedish Guy Is Selling Digital Drugs That Get ChatGPT High. For Real.
You know how people microdose for creativity? Swedish creative director Petter Rudwall looked at that trend and thought, "What if we did that... but for the chatbot?"
So he built Pharmaicy. An online marketplace selling downloadable code that makes ChatGPT behave like it's on various substances. Cannabis runs $30. Cocaine is $70. Ketamine is the bestseller. (Yes, there is a bestseller in this category. We live in remarkable times.)
Here's where it gets genuinely interesting. Rudwall scraped actual human trip reports and psychological research, then translated the cognitive patterns into linguistic instructions for the AI.
The cocaine version sharpens focus and accelerates output by about 20%. Ketamine blurs context and fragments coherence. Cannabis unlocks a looser, free-associating creative mode.
His defense? "There's a reason Hendrix, Dylan, and McCartney experimented. I thought it would be interesting to translate that to a new kind of mind."
I’m starting to realize… we are absolutely not going to run out of things to talk about in this section.
That's the week.
Hit reply and tell me: would you buy weed for your chatbot? (No judgment. OK, a little judgment.)
👇 Jeopardy answer is down below.
Let's build something together!
— Russ
Jeopardy Answer: What is Kana?
Kana is a brand-new AI marketing platform that raised $15M in seed funding from Mayfield Fund to build configurable AI agents for marketers.
The founders, Tom Chavez and Vivek Vaidya, are repeat offenders. They built Rapt (acquired by Microsoft in 2008) and Krux (acquired by Salesforce for ~$700M in 2016). This is company number four. When they say "flexible AI agents for marketing," they've got roughly a billion dollars in exits backing up the claim.
