@_guillecasaus Significantly less ram, but also significantly less development, significantly less features, significantly less security… The ram usage isn’t a bad thing, it’s actually what makes google superior.
@HermesAgentTips Only through iMessage. Hermes manages itself and all my other bots so I dont need more than a “do what I say” and a “heres your response and proof” kind of situation.
11 hours until the @NousResearch hackathon… or I could sleep. Let’s see what happens. I’m cutting it tight. I don’t expect to win using Windows XP, but I hope it inspires people to revive their old tech lying around and learn some things along the way.
Temu sent me a free Chinese camera and it turns out it has a pretty serious security hole. Pretty much anyone could see me, hear me, talk to me and move the camera around 🤪
So I blocked it at my router, and now I'm turning it into a Hermes plugin. Hermes gets to see, hear, talk and look around instead of the whole internet. Spyware to Hermes robot.
And that's not all that's coming. You'll all figure out there's something seriously wrong with me, in a good way
@iamlukethedev Just the recognition makes me smile, feels like they are saying “Okay so you don’t like our harness? Well be set free.” But like in an official kind of way where they are not going to block us from using it our way one day.
Unlimited AI unlocked?
Earlier today, I told Hermes to burn through my remaining ChatGPT usage because my subscription expires tomorrow and I wasn’t sure I’d renew.
Well, it’s been working on HermesXP all day, and it’s STILL going.
I checked the actual usage stats:
~145 million tokens processed (95.7% cached)
1,537 model calls
30 commits pushed to HermesXP today
Grok has been orchestrating the work, with both ChatGPT and Grok handling implementation.
Two things surprised me:
1. The efficiency is ridiculous.
I expected to hit my usage limits pretty quickly. Instead, Hermes has been running for hours, making extensive use of prompt caching while continuously moving development forward.
I’m genuinely considering keeping both subscriptions if this level of performance continues.
2. The reliability is what really got me.
I haven’t had to babysit anything.
Hermes has been managing an open-ended development task, delegating work, reviewing changes, running tests and continuing without constantly asking me what to do next.
Normally, a standalone ChatGPT session would’ve stopped for clarification or waited for another prompt by now.
Hermes just keeps the work moving.
I’m starting to think the real breakthrough isn’t necessarily smarter models. It’s building a harness that can keep them productive.
And yes, Hermes is still working as I write this.
@justsyun@modretro You could have one button where if you held it down, it would reset the orientation so you could hold your hand how ever is more comfortable for you. Really cool idea and cant wait to see where this goes!
Want a new feature for ChatGPT?
Wait 5 minutes and a new Hermes plugin will unlock that. My feed is nothing but Hermes can now do insert_anything_here.
Before I used to have to engineer some crazy hacks to make ChatGPT do stuff, now I just tell Hermes to do it with ChatGPT.
@_Tom_96_@NousResearch No, I think @NousResearch is better than that. But I think your X post traction is good insight to see if your plugin will get selected.
This guy's probably twice my age and I still want to hang out and nerd out over tech with him. I've only been following for a few days and you've already given me way too many ideas, my to-do list hates you 😂 Seriously inspiring though, can't wait to see what you do with the new laptop!
Hermes keeps getting better. At first I was configuring it 24/7, and now I'm mostly just tweaking it for what I need. I can't wait to see how different all our agents are in 5 years. They'll have their own personalities, just in a mechanical way. Some will run businesses while others repair and improve those agents. Hermes feels like the Linux of AI, and that's what makes all of this possible. Now someone just needs to make a Hermes robot body plugin 😂 We'll have them running around before big tech does lol
Made a Hermes plugin so my agent can't just write whatever it wants into its own memory 😂
New memories go into a queue first, and a reviewer agent checks them every so often to approve, reject, or revise them. Anything that touches permissions waits for me to approve it.
hermes plugins install alxspiker/hermes-memory-reviewer --enable
https://t.co/xssVxZ3tXW
Thanks @NousResearch again 😎
This would have been impossible with “frontier” AI companies.
We might need more @NousResearch plugin categories. Suggestions because you didn’t ask: Hardware & IoT, Gaming, Development, Documents, Security, Creative, and Science.