Can't speak for other places, but at Meta everyone effectively has a free GPT4 subscription.

If you’re an eng and not using ChatGPT/copilot you’re going to get left behind pretty soon.

It's unsettling how many times GPT 4.5 has been wrong about something that I can easily see and point out and often verify by finding documentation or testing. I don't know how many times I've asked a question and told the wrong thing.

> It's unsettling how many times GPT 4.5 has been wrong

There is no GPT-4.5. Current OpenAI GPT models include various date and context-size checkpoints of 3.5-turbo, 4, and (in preview) 4-turbo.

That’s too bad for you. I’d say for general corporate development, the output is directly usable 90%+ of the time.

Sure, if you are doing unoriginal low quality work.