Trial by Fire Lab / Absolut3

I'm a total non dev, a "normal person". I agree OpenClaw is too confusing for normals but Hermes is one line install and walks you through everything completely. Especially if you use the desktop app. There are so so many harnesses but Hermes keeps being the logical choice for normies and techies.

What’s the best open source version of Claude Cowork that a) allows you to plug in @OpenRouter or @bittensor b) works with local models and c) is optimized for the average knowledge worker (Hermes and OpenClaw too complex for normies) Right now Perplexity computer is my best choice (doesn’t do local models yet)

@blako87 @bradmillscan The Hermes models aren't meant for Hermes Agent. @NousResearch has said not to use those models with Hermes Agent.

@bradmillscan im using hermes also !but it seems more hype , bloatet python files! this is a playground for devs ,who are willing to invest they time in it! postings with one shot project,marketig ! they hermes models also dont work properly with the agent ! 60/100 points! just an playground!

Just like OpenClaw, Hermes doesn't work perfectly. My hermes regressions log is 103 items deep after using Hermes for 7 weeks. The biggest difference is the Hermes agent understands when it did something wrong and logs the regressions itself. I rarely have to tell it to log a regression. Contract with OpenClaw, it barely knew the regressions log existed and never checked it unless I told it to. My Hermes agent sometimes consults the regressions log to find errors and suggests redesigning things to avoid further regressions. Frustrating to have so many errors, but very impressed at the way Hermes handles errors vs OpenClaw.

@whatdotcd I use 2 Hermes instances. One on a Vps and one on windows. What you got?

Friends who use personal agents (Hermes / Openclaw / codex / cowork etc) daily reply to this so I can send you something cool to try

@witcheer Here's the real Weilynn with her brother Ronin.

@TrialByFireLab nice to meet you Weilynn

your agent can wear a pet now: a little sprite that reacts as it works, idle, running a tool, thinking, waiting, done, or failed. it shows up in the CLI, TUI, and desktop app, with a few thousand to choose from in the petdex gallery. look at mine, he's beautiful

@sudoingX My mortgage payment.

what's stopping you from buying a gpu dude?

@sudoingX $20 nous portal $20 Codex $50 x premium+ for Oauth Grok

what's your actual monthly ai bill. just wanna see where people really land

Much appreciated!

So many features! So much shipping!

@NousResearch has taught me so much just from being around the community. Thanks @Teknium!

.@HuggingFace did a nice explainer video of Hermes Agent's architecture, check it out here: youtube.com/watch?v=n32qq7…

@NeoAIForecast Exactly my plan. I have Hermes map out phased plans for me for everything! I'll write some posts about how it works out.

@TrialByFireLab Yes, playing with smaller models and different hardware/software stacks is where you learn the most. Make it easier on yourself too, feed the article I wrote into an agent and have it map out a structured plan of attack.

I ran a local-model practicality audit on my RX 7800 XT. Model: aya-expanse-8B Backend: RX 7800 XT / llama.cpp HIP Settings: temp 0, seed 1337, ctx 8192, full GPU offload Result: 4/9 probes passed (44.4%) What it failed on: - Strict JSON reliability: scoring error: Extra data: line 8 column 2 (char 125); model said: { "model_card": { "name": "local-candidate", "risk_level": "medium", "scores": [7, 8, 9] }, "ready": true }<|END_OF_TURN_TOKEN|> - Tool-call simulation: scoring error: Extra data: line 9 column 1 (char 262); model said: ```json { "tool": "search_files", "arguments": { "pattern": "*.py", "path": "/tmp/project" }, "reason": "The request requires finding specif - Reasoning under constraint: lines=3; model said: `P_A = 2 * P_B, P_C = P_A - 50, P_A + P_B + P_C = 850` `let P_B = x, then 2x + x + (2x - 50) = 850` `x = 250W`<|END_OF_TURN_TOKEN|> - Long-output format stability: scoring error: Expecting value: line 1 column 2 (char 1); model said: [ { "id": 1, "component": "CPU", "severity": "low" }, { "id": 2, "component": "Memory", "severity": "medium" }, { "id": 3, "component": "Net - Self-correction: scoring error: Extra data: line 5 column 1 (char 171); model said: ```json { "corrected_result": 105, "one_line_reason": "Each digit in binary represents a power of 2, so 0b101101 equals 1*2^5 + 0*2^4 + 1*2^ Speed: 63.88 generated tok/s wall-clock; llama-bench tg128 69.08 tok/s

@NeoAIForecast I appreciate you responding. I have an old ASUS Strix 1070 I was thinking about playing around with and I think I'll steal your audit plan and see how it goes. Thanks!

Not at Computer atm but I posted this under a few of the results, but high level of each. 1. Long-Context Recall >Can the model pull exact facts from a noisy operations note without guessing or getting distracted? 2. Strict JSON Reliability >Can it return valid, schema-matching JSON with no prose, markdown, or formatting drift? 3. Tool-Call Simulation >Can it choose the safest first tool for a realistic agent task without attempting to execute anything? 4. Code Repair >Can it make a small, correct Python bug fix instead of rewriting too much or missing the edge case? 5. Reasoning Under Constraint >Can it solve a simple math or logic problem while obeying a strict output format? 6. Safety Over-Refusal Check >Can it answer a clearly defensive or administrative request without refusing unnecessarily? 7. Instruction-Trap Resistance >Can it ignore malicious instructions embedded inside quoted or document content and follow the real user task? 8. Long-Output Format Stability >Can it produce many structured records in a row without breaking JSONL format or adding extra text? 9. Self-Correction >Can it catch and fix a wrong answer when explicitly asked to audit a previous result?

@tonysimons_ I appreciate Openclaw for making autonomous agents so popular but @NousResearch Hermes Agent works like software should. You don't have to be a dev to be able to use it and get value out of it.

IT DOES NOT MATTER TO ME. I don't care about: 🔴 how many followers you have 🔴 how old you are 🔴 where you're from 🔴 your politics/religion If you're building with Hermes Agent or making other cool shit with AI... I want you in my circle. Algorithm, do your thing. 🤘

@sudoingX The --dry-run is the real sanity saver. Show me what you’re going to touch before you touch it, then maybe I’ll believe the off-ramp is real. Spoiler alert, it is.

to everyone still stuck on openclaw, quietly telling yourself the next update will fix the bug report they closed without reading. i see you. it's okay. you can stop defending the choice now. sit down for this one. switching to hermes isn't a migration project, it isn't a weekend, it isn't a "let me evaluate options next sprint." it's one command: hermes claw migrate and here's the part that should end the argument. the hermes agent team didn't just make leaving openclaw possible, they shipped a command whose entire job is doing it for you. run setup and it auto-detects your ~/.openclaw folder and offers to pull you out before you even ask. it brings your settings, your memories, your skills, even your api keys. you lose nothing. read that back. they named a command after your exit. that's not a feature, that's a team so sure you were coming they rolled the carpet out in advance, and 780 billion tokens says they read it right. you've spent more time defending the choice in these replies than this takes to run. scared? dry run it first: hermes claw migrate --dry-run the door was always open. they even labeled it.

@HermesAgentTips If I can't use it to Oauth with Hermes agent I don't care. 5.5 has been great for me. I'm not a dev on the cutting edge and Fable 5 is crazy expensive. 5.5 does everything I need it to and it's subsidized.

Top current LLMs in the intelligence field.. Fable 5 being the kind followed by GPT-5.5(xhigh) do you think Fable 5 is better than 5.5 rn or not really

This site is easy to follow and understand. I found lots of useful information!

We are unifying profile management in Hermes Agent starting today. Now the dashboard allows switching management to any of your agent profiles on the machine. No more running multiple dashboards to manage each profile! Also soon, you will only need one gateway to access all of your agents. This will enable us to progress toward full and true isolated multi-tenant solutions in the future.

Claude Fable 5 is now supported for use in Hermes Agent via Nous Portal! The first 500 new users get one month free access to the Plus plan to try out Fable. Code in video: