Do you really need an agent?

I'm an engineer. I use AI both for my work and my personal life, basically every single day. I don't think "AI is a fad".

Here's the thing, though: For the past few months "agent" has been the buzzword. 99% of people talking about it have no clue how it works or what it is exactly, but it's the thing.

"Yo, I'm running 39 parallel openclaws in my 45k Mac Mini tower I built at home" "man these are making me so productive, I'm producing so much stuff you can't even begin to comprehend it" "this changes everything."

You've read at least ~50 variants of each of these. You see it all the time, to the point where you're wondering whether maybe you're missing out on it - maybe you really are "not gonna make it". And while thinking this you scroll down and another related post shows up.

This has pissed me off so much my hand has been forced to post here this half rant, half discussion starter: Do you really need a personal agent? Are you really spending so much time every day on the minor tasks it can do, that you actually trust it -unsupervised- to do them for you? Is it really worth the risk letting <current SOTA model> respond to your emails for you? Have you actually thought of how screwed you could be if it just... does the wrong thing?

I really can't stop coming up with questions like these every time I see one more of these posts, one more analysis on what the best set up for your personal agent is today (which is just so much better than it was 5 days ago I swear!).

I see the timeline and can't help but think: there's no way this many people have so much work that can be automated... right? And if they do have so much work that is easily automated - was that work worth doing anyway?

THE ACTUAL QUESTION: Have any of you actually set up a personal agent and kept using it 2 weeks after? Has it actually helped you with your daily tasks in any noticeable way? Am I really not gonna make it by not buying a Mac Mini?

That's it, I'm not promoting anything or building the next agentic startup btw.

4 points | by g_br_l 2 days ago

7 comments

  • runjake 8 hours ago
    I played with OpenClaw and see the value of it and the glimpse it provides of the future, but the major thing it showed me is that I'm just not that connected in life. Not to the point where it provides usefulness to me.

    I do frequently use ChatGPT Voice Mode, so I can see a future where a more frictionless version, where I give corporations access to all my life's most intimate data, becomes useful.

  • todteera 14 hours ago
    The whole go buy a mac mini to run openclaw thing I don't really get. What levels of personal agent automation do people really need? I use claude code every day and I don't think there's anything that couldn't be done with some skills and subagents.
    • g_br_l 13 hours ago
      I honestly don't know. I think one of the reasons openclaw specifically got so popular is because it reached a certain critical mass (in popularity) that made it become sort of the "first contact" with an agent for a certain, less/non technical audience - in the sense that it showed them features that they didn't know could be achieved with other, already existing tools.

      just speculating though.

  • stonefull 2 days ago
    That's a relevant question. But, like any tool ever developed by humans, many will misuse it, and few will take advantage of its resources.
  • maxim_manylov 2 days ago
    I still couldn't figure out what are the purpose of macminis, especially a lot of them? What for?
    • g_br_l 2 days ago
      more agents, I guess
  • moomoo11 11 hours ago
    agents can work with other agents, which is the interesting part

    as for the openclaw obsession, i think eventually -

    1. it will have a small niche market for off the rails experience

    2. apple/google will build a walled garden theme park experience on siri/assistant and it will be pretty good and integrated. 99% of people will use this

    Hey Siri, order my usual at Starbucks and schedule it for 7am.

    It will use whatever APIs are exposed and schedule it. You run late, maybe 15 min before before you realize you'll be late. "Hey Siri, can we push my Starbucks order ahead by 15 minutes?" API says not possible, order already started. "Unfortunately, your order is already being made."

    3-6 months later, agentic sdk integrations will make it possible for the Starbucks app to do close to real time order management based on location.

    Ta da. Actual shit that is useful.

    I think the only apps we'll use will be business apps and social/fun/engagement farming type of apps. Most other things, like ordering food or shopping, will be done by the on device assistants. Most of those are just static APIs and web hooks.

  • scattered_tabs 1 day ago
    [dead]
  • causalzap 18 hours ago
    [flagged]