Did they just update tasks? (positive update)

ChatGPT
, ,
1

I wasn’t using scheduled tasks until now because in my experience the results were inaccurate and often brought outdated poor results, using like a lightweight gpt-4o-mini engine.

Then today I decided to give it another shot, and noticed it now is not making the same mistakes and the results do seem more acurate and are using the new reference quotes (with highlighting).

For example, I asked to schedule a task to show me news from today, and make sure they aren’t just old news being republished, and it worked nice. Before it would bring me 2 month old news that got mentioned in a shady website.

Weird thing is that seemed to have been no news on this, except for a brief mention that we can now create tasks using o3 and o4-mini.

What do you people think, have you tried them recently?

2

Update: some tasks do run online, while others look like they run offline, with no link references and completely outdated information.

The good side is that once working they seem to continue doing so, but this seriously lack objectivity on how can we control wether it can access web or not.