You Can Now “@” GPTs: A Small Update With Big Implications

Credit: DALL-E

It’s an old rule in product releases: Don’t ship anything on a Friday. After all, if anything goes wrong, you don’t want your engineers working over the weekend. Only the most minor of updates should go out at the end of the week.

You’d be forgiven, then, for assuming OpenAI’s launch of the ability to reference GPTs on ChatGPT with a simple “@” character — released late Friday afternoon — as minor. And from a technical standpoint, it probably was: The feature is really just a shortcut that negates the need to either find the GPT you want either in the GPT Store or from your “saved” list in the ChatGPT sidebar. However, the “@” update has a couple of big implications:

? Friction, ? Utility

First, the update cuts down friction — arguably the entire mission of AI in the first place. It’s still early days for GPTs, but many do appear to address specific use cases better than the out-of-the-box ChatGPT experience, but that utility only helps if people actually use them. With the update, you can even decide to do so mid-chat: for example, you could first order up an article on low-calorie lunch ideas with avocados, then ask your favorite copy-editing GPT to give the copy a pass in the next prompt.

The adjustment may be technically trivial, but from a productivity standpoint, it eliminates the manual steps of a) hunting for the GPT, b) launching a new browser tab or chat, and c) copying and pasting between windows. All this amounts to reducing the “mental load” of the task significantly. The GPT Store still has issues with discoverability (it’s hard to find GPTs that are both good and specific to your work), but the “@” update has the potential to greatly increase the usage of the ones that are found. (Now if only GPT developers could make money on them.)

Subscribe now

The Limits of Retail AI

This update, like the GPT Store debut before it (and really everything since the company’s Dev Day in the fall), shows just how much OpenAI is focused on “retail” AI — capitalizing on the popularity of ChatGPT by rapidly iterating on that experience for individuals, locking them into the platform.

The company is following the Apple playbook almost to the letter — put out the best possible customer experience, cultivate a massive user base that evangelizes your product, and wait for your competitors to fall off the longer the race goes on.

The strategy may work, but as interesting as the GPT Store is, it’s still primarily a retail play. If you’re a business, chances are you’re going to want something more custom, plus you’ll have the added concerns about data privacy and locking yourself into a single model. The GPT Store can show you what’s possible, but what you really need is something tailored to your workflows.

Sure, OpenAI offers an API and enterprise solutions for exactly that, but once you’ve decided to go that route, you’re going to look at all the options, and there are many. This is where the Apple strategy redux breaks down: In the world of AI, there’s no hardware lock-in (at least not yet), so companies using it will simply use what works for the least cost.

The strategy might work well for OpenAI: after all, Apple eventually got so influential with its consumer devices that business and enterprise software had to adapt to accommodate the iPhone. But in the world of generative AI, where your competitor is just a browser tab away, OpenAI’s retail focus may inspire the business world to seek out its competitors.

The Media Copilot Meetup is happening in just a few days! If you’ll be in NYC on Thursday, you can RSVP for an evening of stimulating conversation about AI, media, and the future right here. See you soon!

Yes, I Want to Go!

The Media Copilot is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Ready to start using AI like a pro?


Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.