The last Experimental was WAY better at coding + please train for NO PLACEHOLDERS

Gemini API
1

The production version of 1.5 Pro is horrible at coding compared to the last Experimental. Almost not even useable.

Furthermore - for the love of God - train your models NEVER to put placeholders in code they generate without user permission. No matter what I do with the production version, code it writes is littered with placeholders indicating unfinished code, previous code, code that needs to be written… it’s impossible to use … placeholders destroy code

If a user asks for placeholders then fine - but the default should Placeholders = OFF!

3 Likes
2

Another thing I hate about AIs with programming is that it leaves out a ton of stuff and just says “// Previous code here…”

I mainly program on mobile, as I’m generally out of the house quite a lot, and copying and pasting bits manually actually sucks, especially on Apple devices.

If a Google AI employee sees this, please fix both the placeholder and the emptiness of programming responses.

Thx!
Hat

1 Like