expectedwrong hindsight

Google's Beautiful Lie and the Two Prices of AI

The demo costs nothing. The product costs everything. Google forgot to mention the difference.

3 min read 627 words #AI #Google Gemini #product strategy #higher education #demos vs reality
hindsight — nailed it

The beautiful lie was exactly that. Google got caught staging the Gemini demos — edited, not real-time, cherry-picked to the point of fiction. The "two prices of AI" framing held: what it can do in a controlled demo and what it does in your hands are still two different products at two different prices.

Google dropped the Gemini videos this week and they are, genuinely, exactly what everyone who's been thinking about "beyond chat" AI has been hoping to see — multimodal, contextual, something that feels less like a search bar with a personality disorder and more like a system that actually understands what you're trying to do.

The demos are great. Spot on. Go watch them.

Google did forget to mention one thing, which is that none of it exists yet. Won't ship until 2024. The videos are a vision document dressed up as a product announcement, and they are very good at being dressed up.

This would be annoying if it weren't also completely instructive.

There are two prices for AI.

The first price is astonishingly low. You can put together something that produces genuinely impressive results — something that would make a dean's jaw drop, something a client would screenshot and send to their board — with almost no technical infrastructure. Prompt engineering, a capable model, some thoughtful context design. A person who has never written a line of code can do this in an afternoon and the output will look like the future.

The second price is grotesque. Taking that afternoon prototype and turning it into something that actually works — reliably, at scale, integrated with the systems that already exist, maintained across model updates, compliant with whatever legal obligations your institution has accrued over the decades — that price is not ten times higher. It's not even a hundred times higher. It is a different category of problem, priced accordingly.

The gap between those two prices is where the entire AI industry currently lives, and Google's beautiful Gemini videos — which show a future that is real and coming and genuinely exciting, and which are not available for you to use — are a perfect diagram of it.

I think Ball State is figuring this out in real time, which is interesting to watch. The second video looks like what they've been building — or something very close to it — and from what I can tell, the approach matches how the work has actually been happening: person-by-person, iterative, surprisingly effective given the constraints. The first price. The good price.

The problem is that the people paying the first price are now being sold the second price's promise. And the vendors — Salesforce being the specific flavor of disappointment here — have a well-documented strategy of showing up for the sale and then becoming increasingly difficult to reach. This is not a new behavior. It is, in fact, Salesforce's primary product.

My prediction is that the people who are currently frustrated about Salesforce not returning calls will be considerably more frustrated by mid-January, when the gap between what was demoed and what was delivered becomes impossible to ignore.

None of this means Google's videos are dishonest, exactly. The capabilities are real, the direction is right, and 2024 will get there. But there is a particular kind of damage that happens when a great demo creates a procurement cycle — when someone watches a video, gets excited, writes a check to a vendor who promises to deliver that video, and then waits. The check gets cashed immediately. The video remains a video.

The sane move, if you're building in this space right now, is to stay in the first price for as long as possible. Build things that work with the models that exist today. Accept that it won't scale, that it won't integrate, that it will break in ways that require a human to fix. Let the second price come later, when the infrastructure actually catches up to the demos.

Google's vision is where this is going. It's just not where it is.