GPT-6 Astra Ultra test burns 18% of weekly quota in 38 minutes
The author tried GPT-6 Astra Ultra without their usual constraints: it ran for 38 minutes and took the weekly limit from 61% to 43%, with results worse than expected, leading them to argue that task setup matters more than the mode.
I tried GPT-6 Astra "Ultra" for the first time without my usual constraints. It ran for 38 minutes, took my weekly limit from 61% to 43%, and honestly the result was worse than I expected. This is why I keep saying the mode matters less than how you set the task up. Medium, High, Ultra… if the agent doesn’t clearly know what you want, what it should touch, and where it should stop, it can spend a ton of time going in circles and still miss. Give it the context it needs, set clear rules, and let it ask questions before it starts. That usually gets you a better result without wasting usage on stuff you didn't asked for.