The Allbirds Moment
A few weeks ago, Allbirds—the company that sells comfortable shoes made from recycled materials—announced it was now an AI company. The stock price jumped something like 700%. Let that sit for a second. A shoe manufacturer rebranded itself into the AI space and the market responded like someone had just cured cancer.
I spend my days actually integrating AI APIs into products. I work with Claude, GPT-4, open-source models, vector databases. I know what it feels like when an integration works and when it doesn't. And I can tell you with certainty: rebranding is not the same as building.
What We're Actually Seeing
There's a difference between AI as a utility and AI as a narrative device. When I add a recommendation engine to a product, that's AI as utility. When Allbirds slaps "AI" on a press release and the market rewards them with a 700% bump, that's narrative. I'm not even skeptical about this—I think I understand it. Investors got burned by crypto. They need a new story. AI is available.
- Real AI integration means rewriting data pipelines, managing model drift, handling edge cases nobody planned for
- The press release version means one announcement
- Stanford's 2026 AI Index actually shows performance improvements across measurable benchmarks, which is real
- But benchmarks aren't the same as business value, and I'm not even sure they should be
- Most companies adding AI to their stack are doing it because they're afraid of being left behind, not because they solved an actual problem
Here's where I get stuck though. I genuinely don't know if the hype-first approach is wrong. When everyone starts expecting AI features, companies that move fast—even if they're moving fast with mediocre implementations—might win against companies that wait for perfect solutions. Speed might matter more than competence here. I hate that possibility, but I can't dismiss it.
The Inevitable Trap
"AI is inevitable." That's the phrase I hear in every product meeting now. And it does something weird to how we think about building things. It transforms the question from "does this actually help users?" to "how do we make sure we're not left behind?" Those are not the same question.
When inevitability becomes the premise, you stop asking hard questions. You stop asking whether an LLM is actually better than a decision tree for your problem. You stop asking whether your users need personalization or just better search. You just... add AI, announce it, hope the market notices.
I've done this. Not with a shoe company rebrand, but I've definitely shipped AI features that existed primarily to have AI features. A chatbot that could've been a form. A recommendation system that worked slightly worse than the algorithm it replaced but looked more impressive in demos. The pressure is real, and the pressure is everywhere.
What Actually Matters (Maybe)
The Stanford report says AI got better at specific, measurable tasks. That's not nothing. That's actually important. But "AI got better at image generation" is not the same as "image generation solved a real business problem for Allbirds customers." And I don't see evidence that shoe buyers suddenly care about generative features.
What I notice with actual successful AI implementations: they solve specific friction points. They're boring. A company uses AI to auto-tag support tickets—saves 8 hours a week. A team uses Claude to generate boilerplate code patterns—cuts scaffolding time by 30%. Nobody announces these things. Nobody's stock price septuplets.
The inevitable trap is believing that inevitability means universal applicability. That if AI is getting better at tasks, then every company needs it everywhere. That if your competitor mentions AI, you're losing. It's a narrative that benefits venture capitalists and it costs smart engineers a lot of sleepless nights.
Where I Actually Stand
I think we're in a peak hype moment. I think some of that hype is justified—the models really are getting better—and I think most of the market reaction is disconnected from actual utility. I think companies will continue announcing AI products and some of them will create real value and most of them won't and the market will eventually adjust. And then something new will cycle through the same pattern.
What I'm not sure about: whether this cycle is inefficient or just expensive. Whether the failed experiments are waste or necessary exploration. Whether building the boring, specific AI features is the smarter play or the cautious play. Whether I should be shipping more aggressively or more thoughtfully.
Allbirds' stock price might hold. It might crater. But either way, it tells us nothing about whether AI actually works for shoes.
