Steffen Fiedler

@witcheer @stackwalnuts Super relevant to what I'm trying right now on Hermes: Obsidian-backed memory with ongoing cycles so each run closes with structured updates and the next run starts from a fresh state + goals. Keep us posted on week-one outcomes.

trying a new memory tool for my AI agents: @stackwalnuts basically it organises your context into "walnuts" - structured files with their own identity, state, history, and task queue. separate domains that link to each other. your agent reads them at session start, writes to them at session end. set up 4 walnuts on my mac mini yesterday: personal identity + content strategy // day job (CDP protocol, partnership pipeline) // AI agent capabilities // business admin wired my 24/7 cron agent to write to the walnuts automatically. end of every research session > log entry prepended to the right walnut. end of every content draft > captured to the personal walnut. what i'm watching: - whether the agent reliably runs the walnut update at the end of long sessions - how context quality holds up after 50+ entries - setup friction: ALIVE was built for openclaw, I'm running hermes agent. had to write a custom skill + shell script to bridge them. will report back in a week

@CoastalFuturist Yes, interested! Coming for Hermes.

If there’s enough interest I’d like to make a group chat for people using openclaw / hermes agent heavily I really want to understand some good use cases, best practices, and just have a place for people to talk shop Comment if you’re interested

@LottoLabs I'm so regretting buying the 16GB Mac Mini…

Hermes Agent and qwen 3.5 27b on RTX6000 getting a nice 48TPS on vLLM Testing 3090 + vLLM next

@kr0der Hermes is state-of-the-art in agentic systems. Straightforward setup, slim config, never drops a job, and has self-healing capabilities.

gonna give Hermes Agent a try this week, anyone tried it yet? it has the coolest name not gonna lie Hermes

@0x_Vivek @tunahorse21 Any source where I can get more details on that approach?

@tunahorse21 Hermes sync architecture is clean. Real alpha is connecting it with SQLite for persistent memory across sessions. It's how I keep my PR learnings alive. 🦞

okay hermes agent kind of bangs, simplest setup out of all the personal agents ive tried fed it a bunch of repos via telegram and it installed cleanly connected todoist cli with just the repo name

@tunahorse21 Slim config, straight forward setup and never dropping a cron.

@techNmak Why would running a small local model be a good choice if the quality of the resulting source code is junk compared to Opus? Instead, using a local model for housekeeping, data retrieval, and data collection tasks can be great to keep the bills down.

Claude Code can run entirely on your local GPU now. Unsloth AI published the complete guide. The setup itself is straightforward - llama.cpp serves Qwen3.5 or GLM-4.7-Flash, one environment variable redirects Claude Code to localhost. But the guide is valuable because of what it explains beyond the setup: Why local inference feels impossibly slow: Claude Code adds an attribution header that breaks KV caching. Every request recomputes the full context. The fix requires editing settings.json - export doesn't work. Why Qwen3.5 outputs seem off: f16 KV cache degrades accuracy, and it's llama.cpp's default. Multiple reports confirm this. Use q8_0 or bf16 instead. Why responses take forever: Thinking mode is great for reasoning but slow for agentic tasks. The guide shows how to disable it. The proof it all works: Claude Code autonomously fine-tuning a model with Unsloth. Start to finish. No API dependency. Fits on 24GB. RTX 4090, Mac unified memory.

@LottoLabs Same here, Hermes + Gwen 3.5 4B locally for housekeeping and optional 14B for background tasks where tidiness is key.

Little hermes agent running w/ 4b qwen 3.5 running skills to manage your node. This is the way forward.

@Teknium @shawn_pana Whats the difference to Hermes‘ default web access?

@shawn_pana Is that a new browser backend? Can you make a PR?

Hermes agent is insane I gave Hermes access to Browser Use so it has access to my social media accounts. It retains context about my codebase, tone, and workflows. Now i just ask it to post on X for me...

@decisionoracle @Teknium Maintaining a growing configuration became a balancing act with Openclaw. In comparison Hermes Agent has a very slim an thought through setup.

@stffnfdlr @Teknium what do you like better?

We just released Hermes Agent! In my humble opinion a very good blend between coding agents like Claude Code and generalist agents like Clawdbot. Been working on this for the last month or so now - started as a way for us to have agentic primitives for datagen and RL and got inspired by the agentic revolution of late, so been expanding it's scope and capabilities non-stop! Hope you all enjoy.

@Zeneca Fully transitioned to Hermes Agent. In comparison Openclaw felt like a tar pit.

thoughts on openclaw vs hermes? has anyone switched from one to the other, successfully bringing their memory systems and other things?

@usgraphics What do you use to layout the header and table?

Our new backend RFQ system:

@stphnthiel @jackrusher Please keep in mind that queuing is something very natural to the natives in Brandenburg.

Flying out of the new BER airport for the second time. I just want to say publicly that I am very willing to teach free workshops on queuing theory to architects. en.m.wikipedia.org/wiki/Queueing_…

@stphnthiel @laessr My inspiration has been the white lies approach to faking context with artificial background noise. Like music boxes that provide ambient sounds of traffic or bars when you call your actual partner from an hour hotel. Those must have been from the 60s.

I’m working on my first #fictional iphone app. It imitates user behavior based on generated fake data in order to protect users or give them new possibilities. instagram.com/p/BiMBhtmjiyn

@hamoid @jackrusher Homework turns out to be fun again with @quilist – Next stop @thing_clj!

So glad this is happening – thanks @toxi ! twitter.com/thing_clj/stat…

Very much looking forward to: "The Epic Struggle of the Internet of Things" by @bruces out on @StrelkaPress strelka.com/en/press/books…

@jackrusher may the mighty pedal gods bless you and your équipe with furious tail winds! #cyclesafe

@jackrusher @toxi @stphnthiel About to pitch @thidotng as the next theme for our gatherings!