So I think most of it still holds up, but I think what I'm most impressed by is the recent deep research agent from OpenAI. It really shows you up close the true potential of agents. With that said, I've seen a lot of people still claiming there is some inaccuracy, which can obviously get really bad if you are building, for example, financial reports or legal stuff. For these types of agents, the key will be trying to get them to only pull from really authoritative sources.
As for some other parts, I still see those same companies out here claiming to have agents when they frankly suck and can't do anything. Just up to us to filter through all the BS. I'm still worried a lot of people in the general public will fall for these sketchy claims.
But all in all, we should be excited about the potential. It's only February and I'm already blown away to be honest.
Thanks for this thoughtful analysis, Alex. As a historian, I find fascinating parallels between your observations and previous tech transitions. Your point about starting small and building expertise gradually seems especially wise - it's the kind of measured approach that often gets lost in the AI hype cycle. Really appreciate you sharing these insights about the current state of AI agents and their practical limitations.
Happy you found it insightful, Sean. I make quite a few of those parallels as well. Probably owed to the fact that I graduated with a history degree. Always looking at these developments through that lens.
The topic of accuracy and perception is a crucial one, but the way you reported it here is really fascinating and surprising. Especially in terms of implications for high-risk decisions.
Probably the passage I appreciated the most in this issue is related to experimentation: how this uncertain timeline on developments (especially this year) can be taken as a great way to take the time to see what AI can do, how to use it, where it can already be effectively used. Thanks for elaborating this reflection!
Hi Alex, any updates to your thoughts since the current releases?
So I think most of it still holds up, but I think what I'm most impressed by is the recent deep research agent from OpenAI. It really shows you up close the true potential of agents. With that said, I've seen a lot of people still claiming there is some inaccuracy, which can obviously get really bad if you are building, for example, financial reports or legal stuff. For these types of agents, the key will be trying to get them to only pull from really authoritative sources.
As for some other parts, I still see those same companies out here claiming to have agents when they frankly suck and can't do anything. Just up to us to filter through all the BS. I'm still worried a lot of people in the general public will fall for these sketchy claims.
But all in all, we should be excited about the potential. It's only February and I'm already blown away to be honest.
Thanks for this thoughtful analysis, Alex. As a historian, I find fascinating parallels between your observations and previous tech transitions. Your point about starting small and building expertise gradually seems especially wise - it's the kind of measured approach that often gets lost in the AI hype cycle. Really appreciate you sharing these insights about the current state of AI agents and their practical limitations.
Happy you found it insightful, Sean. I make quite a few of those parallels as well. Probably owed to the fact that I graduated with a history degree. Always looking at these developments through that lens.
The topic of accuracy and perception is a crucial one, but the way you reported it here is really fascinating and surprising. Especially in terms of implications for high-risk decisions.
Probably the passage I appreciated the most in this issue is related to experimentation: how this uncertain timeline on developments (especially this year) can be taken as a great way to take the time to see what AI can do, how to use it, where it can already be effectively used. Thanks for elaborating this reflection!