We were warned about AGI, and we got Shopify checkout in ChatGPT

There's so much going on in tech and AI these days that it's hard to keep up. Between OpenAI pushing Pulse, Sora 2, and in-app checkout, and Meta pushing Vibes, there's a ton to unpack. I gave it a few days to see what the standard commentary across my circles was, and here's what I heard and some of my thoughts:

OpenAI Pulse

Reactions to Pulse ranged from "This is revolutionary and I've been waiting for this" to "I'd never give OpenAI this much data… I don't need Sam Altman to sell me flour after recommending I make my grandma's secret recipe this weekend because I'm feeling sad." Realistically, this functionality was always on the product roadmap, but the prioritization of features and releases is always interesting to me.

Perplexity also has a Discover tab that's been sitting around uninspiringly, not doing much but showing summaries around curated topics (at some point, they'll do something with this). Google has always had Google Discover. Everything OpenAI does will be an interesting foil to what Apple could have done with Siri. All of these, minus Siri, are slightly different approaches to a modern-day AI-powered 'personal feed' that isn't based on the friend graph (Meta is old news).

OpenAI, with something like 800 million users, can define a new experience of what it looks like to have AI integrated deep into our day-to-day lives in a proactive way, shifting away from just a need-based chat interface. How will consumers react?

Sora (OpenAI) and Vibes (Meta)

Sora 2 and Vibes sit at the discovery layer of what I'll later argue is a collapsing funnel. They're essentially TikTok competitors where all content generation is AI-powered. The reactions have been predictably split: some see innovation, but many see brainrot and the further decline of our society and young people's attention spans.

Someone on Twitter asked what the most generous interpretation is for why Meta is creating a dystopian app that further removes us from reality. The best response was that it might be a way to keep AI-generated content quarantined from core platforms like Facebook and Instagram, at least for now. I read another charitable take that we're shifting away from consumption being our expression to the creation of AI content becoming a new cornerstone of our digital identities.

I'm not ready to explore the positive use cases here, and I say that as an advocate and proponent of technology (it's getting lonely fighting for tech in my group chats). These products feel like the next iteration of Character.AI—the segment of AI where people get addicted to talking with fictional characters, which can't be healthy. I know there's more to come in this category and that it connects to Meta's long-standing VR ambitions, but the intermediary steps feel troubling. If my AI-generated videos start serving me McDonald's ads, I'm not sure I can handle how dystopian the present will have become.

More importantly, the further we go into AI video, the more we heighten issues around deepfakes and what's real versus altered. The severity of these concerns, compounded by the fact that these products are primarily used by young people, gives me pause. I'm not denying the technical capabilities or even some potential upsides—I'm just not convinced the trade-offs are worth it.

Agentic Commerce on ChatGPT

This is the most interesting and most hyped update, though this was an obvious next step on the roadmap.

I had a conversation with a friend about the complexity of payment processing when AI chat is involved. In essence, unless there's a specific 'checkout' button in the chat UI, payment processors need further confirmation of the user's intent to take on the intermediary money transfer risk. What OpenAI has built isn't true agentic commerce—it's just moving the checkout button into the LLM interface. I'm smart enough to admit I don't fully grasp the nuances here; even after the sixth time someone has explained the payments ecosystem to me, I only retain it for about a day before I forget the twelve players involved in buying my morning coffee.

But here's what I do understand: even users with high purchase intent face friction that prevents in-app conversion. They might want to watch YouTube reviews first, check Reddit threads, use a specific credit card for rewards, find a better discount through a community they're part of, or simply take time before committing to a significant purchase. This complexity is exactly why charging for ads can be lower risk and more profitable than trying to capture the actual transaction. Is this feature as revolutionary as the forty LinkedIn posts on my feed suggest? I'm not convinced.

Why I'm Skeptical About AI Commerce

There's a chart that was floating around a few weeks ago on how consumers are actually using AI, and a large chunk of it is getting practical advice and creating assets and artifacts. Even the 25% of users using AI for information-seeking represents a different experience than Google's ten blue links—it's more conversational, with longer queries. Of this 25%, some portion might be related to product research, and some portion of those might lead to purchases.

But here's where consumer behavior gets complicated. The internet is filled with analysis on why Amazon has never succeeded with luxury goods. When I want to buy plastic spoons, I want cheap and fast. But when I want to buy an expensive jacket, I need to do more research before committing to the highest-reviewed item, which I know can be gamed. You want to game the plastic spoons I'll buy? Fine. But not a $500 jacket.

That's when I need to see what real people are saying. I'll even consider what the big affiliate sites recommend because at least the reputable ones have a framework for how they compare and judge products—commissions are just one component. But more than that, I want to know what a snobby YouTuber has to say. If I bought that jacket after this supposed snobby YouTuber recommended it and it turned out to be terrible and overpriced, if I met the guy, I'd tell him he stinks! Where do I email Forbes to tell them they stink if they give me a bad recommendation? And the brand themselves? Forget about it.

So OpenAI is going to give me product recommendations based on a variety of sources (at the expense of the publishers that created that content) and then channel me to a Shopify-based vendor so they can make a 2% commission. But I'm still going to check out the snobbiest of influencers so that I can feel better about being more responsible with my purchases. That activation energy—the time spent watching videos, reading reviews, engaging with real people's opinions—might be the equivalent of spending time at the mall. It's part of the experience before you transact, especially for higher-ticket purchases.

I know I'm being somewhat facetious here. OpenAI will inevitably expand coverage and try to be thoughtful about recommendations. But this friction has always existed in any commission-based payment model, and I don't see how moving the checkout button fundamentally changes consumer behavior for anything beyond commodity purchases—in which case, maybe they should have integrated with Amazon.

I’d be remiss to not mention DayDream, which has raised $50m to try and solve discovery but singularly focused on fashion and clothing. I appreciate the more precise vision, and more utility focus, so I’ll reserve judgement and continue to follow their journey. Two other players in the space is a Chinese app called Gensmu (raised $60M) and an Indian app called Glance (raised $390 for the parent company). Glance’s AI app is hilarious — it gives you an endless feed of you in different environments, clothes and products. I met someone from the company, who opened the app and I saw an generated image of him rock climbing, and then next one sitting in office, with links to rock climbing gear, then office wear next. It was hard to take seriously.

Two Competing Theories

The most interesting takes I've read about OpenAI's commerce push are these:

The first is that the real play is and always will be ads, not actual commerce and transactions. Integrating commerce functionality might allow them to gather more data, which will eventually feed into ad personalization—and that's where they might actually make money. This seems reasonable. Look at the biggest advertising companies of our time: Google, Meta, and Amazon—all built on CPMs and CPCs. Why commit to capturing sales when you can sell clicks? Even TikTok's affiliate component is small compared to their ad product, but kudos to them for trying. I'm all for people putting affiliate on the map.

The second take is that none of this matters. OpenAI is interested in AGI and needs to subsidize costs until further technical breakthroughs, so they'll build ad and commerce products in the meantime, but it will never be the primary focus. This one is funny because it reminds me of the classic "we were promised flying cars and we got 140 characters." Maybe the next iteration will be "We were warned about AGI, and we got Shopify checkout in ChatGPT" ten years from now when we haven't hit AGI. (This is not what I think will happen, but it’s a funny line nonetheless.)

The Collapse of the Funnel

This post is getting long, but this is a topic I've been interested in for a long time. I might be an accidental affiliate marketing agency CEO, but it's no accident that our clients are household tech companies.

Ultimately, the larger narrative here is about the collapse of the funnel. We talk about full-funnel marketing—getting in front of users, being there when they're deciding, and pushing them to transact. But over the last few years, the lines between these activities have been blurring. From measurement solutions for podcasts, billboards, and connected TV based on identity graphs tracking whether someone engaged and made a purchase, to mainstream publications entering the affiliate space in hordes over the last five years—the funnel has been collapsing for a while.

But the real collapse happens when someone crafts an experience that truly meets consumers across the funnel in a coherent way. My bet is that OpenAI is not that company, or won’t be the main winner. They're combining different parts of the journey, but trust in commerce increasingly comes from people, not platforms. We may need engagement with real people integrated into a single platform to enable something that can truly sit across these parts of the funnel. TikTok, with creator content tied to in-platform commerce, might be the closest example we have today of what this could look like.

I have many more theories on what the winner in this space could look like, but that might be my next startup, so I'll hold off. DM if you want to write a check though.

What This Means for Performance Marketers

In the short term, and since this newsletter is read by thousands of performance marketers, my point of view remains the same: make sure your affiliate agency is smart. It's too expensive to have a mediocre affiliate agency when this space is rapidly evolving and there's near-zero chance that affiliate content won't be a cornerstone of LLM training data.

Finally, if you're not working with creators, get on it (my next newsletter). Similar to affiliate content, creator content is becoming part of the training data that powers these systems. There's even a world where ChatGPT evolves so that users start engaging and shopping, but not with blind trust in recommendations—they'll use them as a starting point, open a creator's YouTube video in-app, watch it, then return to make the purchase all within the same platform. It's hard to imagine high-value searches working without a creator component. It's why we have reviews on Amazon and Google Maps, and why Reddit, Yelp, and Glassdoor exist. The most human thing about us is that we care what other human beings think and have to say.