Generative AI that drafts fast and still gets checked

Generative models are good at producing a plausible first version of almost anything. That's the useful part and the risky part. A draft that reads well can still carry a made-up figure, a quote nobody said or a citation that doesn't exist.

Here we cover new models when they change what a working professional can do, not every launch. We look at prompting that holds up across repeated runs, at accuracy limits, and at where the review step belongs. The rule we apply: if the output will be read by a client, a customer or a court, a person checks it first.

Feed headlines below come from outside publishers and link to their sites. We don't treat an announcement as evidence a model works for your task.

Next on the list: getting started with generative AI, and beginner advice that skips the hype.