Have you verified that any of its answers are actually good? Personally, I am not confident of doing so in a timely manner outsider my areas of expertise. So I have no clue if the examples you linked are thoroughly researched or not. Especially the Israel/Gaza one. That's an adversarial information environment if I've ever seen one. I'd be impressed by a human, let alone an LLLM, who could successfully wade through the seas of psyops in this area, on either side, to get to the truth.
I've had a lot of success with getting ChatGPT [1] to do thorough research with the following prompt combined with Deep Research:
X=[claim]
Do a deep dive into X. Tell me the steelman arguments in favor of X.
Then tell me the steelman counter-arguments to X.
Then tell me the steelman counter-counter-arguments to X.
Then tell me the steelman counter-counter-counter-arguments to X
Think like a rationalist in the LessWrong style. Think with probabillties, using Bayesian reasoning, looking for discomfirming evidence, think critically and skeptically, problem-solve around potential obstacles.
Basically I get ChatGPT to play steelman solitaire.
Here are some examples I've done to give you a flavor:
Starting with 4o, then o3, and now 5-pro as of this writing