Most people still use LLMs like a fancier Google, ask a question, get an answer, move on. But there's a massive gap between that and what's possible now.
If you're running agentic tools like OpenCode or personal agents like OpenClaw, where something is always working for you in the background, you're operating in a completely different league. Not asking questions. Delegating work.
The divide isn't about which model you use. It's about whether you see AI as a search bar or a workforce.
And many people in this space have been building such as this with this mindset for over a year now. The time to catch up is shrinking. You need to build something today.
Replies (22)
AI psychosis for everyone!
Reading this creates FOMO 😂
But as a bitcoiner i understand to distinguish between signal and noise. I have nothing against using AI, the way you frame it though, is noise.
This message comes with love, lets live in a world where we don’t constantly have the need to chase something…
What are some examples of the types of work tasks agentic AI can do where AI hallucinations either aren't a problem, or where the output is something the person coordinating the openClaw crowd can review and exercise final judgment as to the veracity of the output?
There are too many times when I try to use AI tools for production where the output quality is too low or flatly false.
AI working for you in the background is analogous to giving your credit card to your kids to go shopping, while you are snoozing at home in the sun 😂
This, but in markdown
Yeah, the autonomous stuff just goes off the rails frequently still today so I stick to vibe coding. Haven't had time to build the autonomous system I want, but if I had the time, it truly would 10x-100x productivity.
He's not wrong to try try to motivate people to start playing tho, I think probably Hermes or vibe kanban are the best options atm, but I haven't looked at open claw ina couple months.
Hermes, which is supposed to have long-term memory, keeps forgetting basic information all the time, replacing it with bullshit very confidently. Imagine, it also runs autonomously. You're screwed.
In a simple question/answer chat with projects, you can at least keep a custom system prompt as a summary and it will always remembers things at least pertaining to that project.
So autonomous agents are a bit farther away than we thought, regardless of the hype.
Any prompt with engineered context like a book series, courses, app and game prototyping. Etc. yes and mindset.
I haven't played with Hermes yet, but if it's similar to OpenClaw in its memory structure, you need to setup embeddings to make memory work properly.
You can ask your agent to help you set this up for them.
As a competitor to OpenClaw, Hermes positions itself as an agent that remembers everything about you and gets better with time. After a month of using it, I found it to be no better at all. Perhaps, you need to spend a lot of energy fine-tuning its modus operandi, but I have better things to do as much as I like tinkering with stuff like that.
OpenClaw remembers everything and gets better with time too, but only with embeddings enabled.
There are also other options like Memory Palace or Obsidian.
A quick, informed conversation with your agent may identify the problem, if you don't want to abandon your agent with their amnesia 😂
Its beautiful to motivate!
The issue i had was the framing: do it now, later will be too late.
Same what many people say about btc. Buy now, later will be too late. Which is toxic as well. People miss out on life because they are afraid of missing out in btc
This is making me reconsider the idea then.
Because like hell would I ever give a CREDIT CARD to a kid, mine or not. That's giving a blank check. At most you give a stipend , that way if/when it eventually goes wrong, you only lose what you set aside.
Like with your example, the kid only comes home with 20$ of candy and energy drinks, not however much they decide to spend 😅
Can't give free reign to anyone that doesn't really get consequences
This is one of those things that sounds easy when you say it fast.
The reality of these agents is that they're like really unruly employees that will work really hard for a long time, sometimes accomplishing very little. Sometimes they do awesome stuff, but it happens pretty fast. If you give them something that's going to take 24 hours, there's a really good chance that they're just going to burn tokens and accomplish very little.
It's like managing employees, but in a lot of ways it's kind of worse.
I like the energy and focus, particularly for coding/building. My two pents as someone adjacent, but not within the tech space directly:
I fight with it. I use it carefully to spark learning, critique my own positions—seriously, by anonymizing my own and asking it to steel-man them—and to explore new topics.
It can be an “upgraded” intellectual search tool. Not just a “Google” or search bar, but an interrogative interface. It does take careful and serious prompting to avoid just convincing yourself of your priors.
For someone like me—adjacent, not a coder—what am I missing from your post? What’s the core value-add I could be tapping into?
This is great! Shakespeare is already doing at least half of this.
genuine question, what do you spend a month on this, is it tokens or a subscription?
Setup Hermes yesterday, are there some cool skills for nostr?
I use a Z.ai subscription for my OpenClaw and I use a Claude Code subscription and I use OpenRouter tokens with Open Code via work.
It does!
@hzrd149 recently gave a short demo on this at Nostr Nights!
Nostr.boutique &
@Clara are at about 95% although she recently seems to hallucinate more again. Lobotomy time 🧠
