OpenAI Does Not Trust Its Own Models, and Rightly So
OpenAI does not trust its own models, and after the last few weeks I think it is right. On September 28 it…
Everyone uses AI to market their stuff now. It writes the cold emails, the LinkedIn posts and the ads. It even writes the comments under other people's posts. But nobody is listening anymore. There is far more of it than anyone can read. And the people it is meant for are busy with their own AI projects.
I recently wrote about a developer who builds a booking app for hair salons over a weekend. Marketing the app with AI is now as easy as building it. An AI agent finds 2,000 salons and writes each owner an email that mentions their salon by name. It also posts on Instagram and LinkedIn each day. Setting it up takes an afternoon.
The salon owner gets that email along with a dozen others that week. Other booking apps write too. So do agencies that want to run the salon's Instagram, and startups that sell an AI phone receptionist. They all use the salon's name and say something nice about its photos. The owner probably does not bother reading these emails anymore.
Social media looks the same. Pangram, a company that makes AI detectors, scanned a million posts. On LinkedIn, more than 40% of the long posts were written entirely by AI. LinkedIn comments were slightly more likely to be AI than the posts. So if you post on LinkedIn, many of the comments you get back are written by AI too.
The other problem is that everyone is busy. With AI, you can probably do what used to be an eight-hour day in two hours. But that doesn't mean you have six hours left. I think people try to do everything at once now. They still fill all eight hours, and a lot of them work even more. There is always one more thing they can start with AI.
Researchers from UC Berkeley spent eight months studying a US tech company with about 200 employees. The company didn't require anyone to use AI. People still worked faster and took on more tasks. They worked more hours of the day, often without being asked. Many prompted AI at lunch or in meetings. One engineer told them, "You had thought that maybe, oh, because you could be more productive with AI, then you save some time, you can work less. But then really, you don't work less. You just work the same amount or even more."
The salon owner does the same. They use ChatGPT for their Instagram captions, their replies to Google reviews and a new price list. Maybe they are building their own booking page. They have no time left to read or answer what other people made with AI.
Give Vroni a GitHub issue, bug report, spec, or rough idea. It reads the repo, plans the change, writes code, runs checks, and works toward a review-ready pull request.
Take a look at vroni.com