CORRECTION: ChatGPT 4o Mini is not more expensive than ChatGPT 3.5

I had a mix-up in today's newsletter. GPT 4o Mini does not cost more than GPT 3.5. ChatGPT 4o (without the "Mini") is the model that is more expensive.

Draftsmith is using ChatGPT 4o Mini, which is less expensive than ChatGPT 3.5 (which is what they were using before).

So to be clear: Draftsmith 2.0 is presumably using less energy than it did before while still getting higher accuracy. 

My line of reasoning about how the models don't seem to be getting more efficient was wrong. It's much more complicated. Companies are building models with different purposes, and some are getting more expensive and others are getting cheaper. 

Finally ... ugh!!! I hate all these strangely numbered product names.

I had to upgrade my Beehiiv account to send out a correction so soon, so the silver lining is that now I have access to polls!