[jermberj]

> The most common category is that the models make wrong assumptions on your behalf and just run along with them without checking. They also don't manage their confusion, they don't seek clarifications, they don't surface inconsistencies, they don't present tradeoffs, they don't push back when they should, and they are still a little too sycophantic.

Does this not undercut everything going on here. Like, what?

[awsanswers]

It's predictable so you run defense around it with prompting, validation and model tuning. It generates volumes of working code in seconds from natural language prompts so it's extremely business efficient. We're talking about tools that generate correct code to 95% of a solution, the follow up human and automated test review, and second coding pass to fix the 5% are a non issue.