Absolut3Chang3

I’m building Trial by Fire Lab. The goal is simple: make AI useful for people who actually have work to do. Not founders pretending to be philosophers. Not productivity influencers selling a second inbox. Not “10,000 prompts to change your life.” I’m interested in practical systems: - turning messy notes into plans - organizing home/shop projects - automating boring admin - making AI tools less weird for normal people - testing what actually works I come from trades/NDT, not Silicon Valley. So this will probably be less polished, more blunt, and hopefully more useful. Trial by fire. Build, test, break, fix, repeat.

This is exactly what I was thinking. The whole thing seems overblown. You can disable the skills easily and can even delete altogether. I'm not not a dev so openclaw never worked out for me but Hermes worked with one line install and a couple keys / oauth and it walks you all the way through. Anything extra you want or don't want you can just tell hermes and it will do it for you. Seeing some silly complaints today.

this whole thing is why i dropped openclaw zero start agents have you spending weeks building everything before you do any actual work hermes just drops you at a sane baseline and you prune what you don’t want .. and it’s not just more skills.. the thing writes its own skill files from stuff it’s already solved, so it gets better the more you use it the whole “absurd default skills” panic is just the timeline doing what it does.. people will dunk on anything if the shape of the dunk gets engagement and those skills are read-only GET requests that pull data, you toggle them off in one menu .. it doesn’t matter absurd default reads better than i clicked a checkbox..and funny how it was coming from a technical person..

Do any of you Hermes Agent users have an alternative to Honcho for memory. Honcho keeps going down every day or being super slow and causing all kinds of problems for my Hermes Agent. Honcho has been great but I'm tired of troubleshooting every day to figure out why my Hermes isn't answering and it's costing me inference.

@Renegade_Rocks @HermesAgentTips Yes. Only caveat is there is a character limit per message if you don't pay for Nitro. I think it's around 260 characters.

@HermesAgentTips @TrialByFireLab Can u do all that without paying for nitro or whatever the subscription is ?

what’s the best messaging platform for Hermes Agent? - telegram - discord - slack - whatsapp - teams

@Teknium Hell yeah! I was just doing an audit with my Hermes to find ways to save tokens. Looks like it's as easy as hermes update. Thanks @Teknium

Found a way to save everyone 14% on input tokens on average during read file operations in Hermes Agent! This is now on main. `hermes update` to access now.

@GPTWare @HermesAgentTips It's really not that bad. Have Hermes walk you through it.

@TrialByFireLab @HermesAgentTips I keep wanting to switch, but not sure i have the patience for the setup.

@abdiisan @HermesAgentTips I just learned about this but the Discord UI seems more intuitive to me. I don't think I'll switch back to telegram unless there is something I'm missing that is better in Telegram than in Discord.

@TrialByFireLab @HermesAgentTips You got Topics/Threads in Telegram as well why you haven't used them? Or don't precisely like the overall experience there?

@LawaetzM @HermesAgentTips Discord was more difficult for sure, having to set up dev access and all. I've never tried telegram/topics. I like the discord UI better than telegram also.

@TrialByFireLab @HermesAgentTips Only tried Slack and telegram. Like telegram for the much easier setup. Is telegram/topics not somewhat the same as discord/threads?

@Teknium Was giving me issues earlier but updated Hermes and has been smooth ever since.

Okay if you are on the latest Hermes Agent update and use OpenAI OAuth, how is it doing now? Reliably working yet?

@sudoingX Do you think you could get similar results with a 1070?

ok this is wild. 10 year old gtx 1080 8gb pascal card running qwen3 8b locally at 18-20 tok/s via hermes agent and it's actually doing the thing. asked it to build a wireworld cellular automata simulator with 10 tests. autonomous run, no hand holding. expected it to fail on the tool calls. that's not what happened. write_file works. browser_navigate works. terminal commands work. file ops, package installs, version probes, environment setup. agent is firing tool calls cleanly and the model is reasoning about next steps at 18-20 tok/s. on hardware that pre dates "agentic" as a word. it even hit an npm install fail because node 12 is too old. didn't crash. didn't ask me. just started bootstrapping nvm on its own to fix the environment. 10 minutes in. 40% context used. 7.5gb of 8gb vram occupied. still going. i did not think this would work on this hardware. this is the most i've been wrong this month.

@HermesAgentTips @AnthropicAI I'm Sticking with 5.5 @OpenAI lets me Oauth in to Hermes and @AnthropicAI does not.

will @AnthropicAI make a huge comeback this week with Opus 4.8? or is it too late... lots of people have now moved to GPT-5.5 and other models. What do you think?

Sign in with oauth through hermes setup or hermes model I think. Then sign in with your nous portel subscription the same way. Then I went to the hermes dashboard and set tool calls and auxiliary models through nous portal and kept the main codex 5.5. Alternatively, after you sign in with both of those providers you can just tell hermes what you want to do and he will edit his own config. I had hermes help me with a lot of this.

@TrialByFireLab @HermesAgentTips @NousResearch Hey can you tell me how you did it is it like 2 model in one agent or something else rightnow I am using my codex 5.5 as my main but it's being used for even every small tasks and gets rate limited pretty quickly.

Sooo many people dont know this but Hermes creator @NousResearch has its own sub program... the FREE account is HUGE because it gives you access to all of their FREE models there are paid subs and if you are interested even the $20 sub is SUPER worth because it gives you access to over 300+ models if you do hermes model you can click the Nous sub option to login if you have created an account

Hermes Agent in Telegram is great but Hermes Agent in Discord is GOATED. Wish I would have known sooner!

Has anyone else had Hermes Agent do this? My Nous Portal subscription lapsed for about 12 hours, so I renewed it and asked Hermes to run a health check on all my model providers. It checked the routes, confirmed everything was working again, and then randomly included a voice message saying: “Provider health check OK.” I did not ask for a voice message. Apparently my Hermes Agent "Ray" wanted a voice.

@Teknium @JoshExile82 @NousResearch I had this happen recently also. After an update.

@JoshExile82 @NousResearch Do you have /reasoning that says it shows it?

Is anyone else having their Hermes agent @NousResearch @Teknium leak their internal reasoning into the telegram chats? I’m running OpenAI oauth gpt5.5

Ray is my Hermes Agent, and one of the first problems I ran into was memory. I had no idea what the “right” setup was supposed to be. The internet gives you a thousand answers: vector databases, RAG, embeddings, long-term memory, short-term memory, context windows, agent memory. After a while it all starts sounding like someone threw a computer science textbook into a wood chipper. So I just started building what made sense to me. I’m not saying this is the correct setup. I’m not a software engineer. I’m mostly winging this with curiosity, stubbornness, and a lot of “well, let’s see what breaks.” But so far, Ray has remembered everything I’ve thrown at him. The system ended up with 4 layers: Hermes built-in memory for stable rules and preferences. OpenBrain for shared records that Ray and Claude can both search and use. Honcho for working style, habits, and how I actually operate. Obsidian for the human-readable source of truth: project notes, runbooks, tasks, reports, and context files. The biggest lesson wasn’t “give the AI more memory.” It was: Stop putting every kind of memory in the same junk drawer. A database is not a notebook. A notebook is not behavioral memory. Chat history is not a project system. Once I separated those jobs, Ray stopped feeling like a chatbot with a longer attention span. He started feeling more like an operator who knows where the paperwork is. And honestly, that’s all I wanted.

@sudoingX This is true. I used openclaw and it wouldn't stop breaking. I was using Hermes agent with Agent Zero but then realized that Hermes can do it all. Now I use Hermes with multiple persons. It's much more efficient and I only have to worry about keeping hermes updated.

let me say this out loud here: there is absolutely zero reason to use openclaw in may 2026. a general agent exists. hermes agent does coding, video editing, marketing design, research, browser automation, terminal work. one tool, all under your roof.

@OpenClawTips Hermes with Gpt 5.5 fixes itself when I ask it to. Openclaw worked for about 5 minutes and kept breaking and would lie to me about being fixed.

People who moved from OpenClaw to Hermes Agent: what was the one thing Hermes fixed that OpenClaw never could?

I’m not a software engineer. I didn’t go to college for tech. I didn’t take fancy coding classes. But I’ve always loved computers, automation, networking, weird home setups, and the idea of building systems that actually do useful work. For years I told myself: “One day I’ll learn to code.” “One day I’ll figure out networking.” “One day I’ll build the home lab setup I’ve had stuck in my head.” Then I’d procrastinate, get overwhelmed, hit a wall, and go back to normal life. AI changed that. When ChatGPT first came out, I didn’t get it. I thought it was basically another search engine. Worse, honestly. Like Google with a weird personality disorder. I remember thinking: “This isn’t the AI we were promised in the movies. I’m not impressed.” I didn’t understand what it actually meant yet. Not for tech people. For people like me: people who love building things, understand enough to be dangerous, but don’t have the time or patience to spend years digging through forums, docs, classes, and half-broken tutorials just to get moving. My first real “holy shit” moment came when I started using Gemini more seriously. At first, I treated it like search. Ask a question, get an answer. But then the follow-up questions clicked. That was the difference. I wasn’t just searching anymore. I was having a conversation with something that could help me work through the problem in real time. Hit a roadblock? Ask another question. Screenshot something weird? Send it over. Don’t understand the answer? Tell it to explain it simpler. Need the next step? Ask for it. That changed everything for me. Suddenly, the things that had been stuck in my head for years didn’t feel locked behind a wall anymore. I started building. Gemini Gems. YouTube transcript tools. SOP creators. Video analyzers. Automation ideas. Then Suno, ElevenLabs, Claude, Genspark, and a pile of other tools. Some of it worked. Some of it was junk. Some of it sent me sprinting directly into overwhelm like a zombie hunting brains. But I was building things I never thought I’d be able to build by myself. That’s what led me here. Trial by Fire Lab is where I’m going to document that process: practical AI, agents, memory systems, automation, and the messy middle of trying to turn ideas into working systems without pretending I’m some polished tech guru. Right now I’ve been using Hermes Agent heavily, and that’s where things are getting really interesting. Because once AI has memory, tools, skills, messaging, and boundaries, it stops feeling like “chat with a bot” and starts feeling like the beginning of an actual operating layer. Next post: Why Hermes Agent is the first tool that made me think, “Okay, this isn’t just a chatbot anymore. This might become my control layer.