Optivaize

@NFTVortexLab Connected, brother!

@Optivaize yo, your dms are closed, dm me pls

Hermes agent buries OpenClaw and it's not even close. I ran OpenClaw for months. Looked sharp in demos. Then I tried using it for actual work. What OpenClaw kept doing: → Crashing mid-task and losing the thread → Solid in the browser, useless almost everywhere else → Freezing the second I tried to ask it something while it was running → Eating more of my day on restarts than on real output Then I tried Hermes. → Runs on MCP servers, so it plugs into almost anything → I can interrupt it mid-task and it keeps going → Every tool call shows up live, I see exactly what it's doing → Tells me when it's struggling instead of just failing silently → I have not opened OpenClaw once since I switched The features matter less than the fact that I trust it. That is what kept me using it. So we built our whole operation around it. Hermes runs almost everything inside Optivaize now: → It powers Malvin, my personal AI assistant. After every meeting, my Plaud recorder sends the notes to Malvin. He turns them into a PDF, drops action points in my CRM, and sends follow-up emails to the people I met. The emails come from Malvin, not me. I am just in CC. → It runs our internal Slack bot. Anyone in the team can ask it to pull data, draft a follow-up, update Teamleader, or kick off a workflow. 90% of our team uses it weekly. → We built a custom MCP server for Teamleader so Hermes can read and write to our CRM the same way it talks to Slack or Drive. → Plugged into Composio it can reach 10,000+ tools without us writing a new integration every time. It is not a chatbot we play with. It is the operational backbone of the company. Big shout-out to the Nous Research team. Shipping something this stable in agent-land right now is genuinely rare. What is the AI tool you keep crawling back to even though it keeps breaking on you?

@tpritha03 ECC is one of those repos we keep meaning to test seriously and never make time for honestly. the agentshield piece especially. we run a similar harness pattern with hermes + custom mcps and the gaps are usually in the security/observability layer. what got you into it?

Stumbled across Everything Claude Code today and this might be one of the most ambitious open-source AI engineering repos I’ve seen in a while. Most repos give you prompts. Most frameworks give you agents. This thing feels like someone tried to package an entire AI engineering operating system 😭 - 30+ specialized agents - 60+ built-in skills - Planning, TDD, reviews, debugging, refactoring workflows - AgentShield security testing with 1,282 integrated tests - Memory systems, hooks, commands, orchestration setups - Support for Claude Code, Cursor, Codex CLI, OpenCode, etc. Getting value from coding agents isn’t just about using a stronger model. It’s about workflows, memory, delegation, context management, security, planning, and repeatable systems around the model. This repo is trying to solve exactly that. Definitely testing parts of it on future projects.

@delancbeb Let's connect!

looking to connect people on X if you're into - building SaaS - vibe coding - AI tools - Building your X account - shipping in public figuring it out as you go.. Say Hi or drop what you're working on looking to connect with active ones 👋

@NikitaKolyadin @X Let's connect!

Hey @X Algorithm I'm looking to #connect with builders in: 🤖 AI agents & automation 📷 Backend / APIs 📷 Solutions architecture 📷 Solo & indie founders 📷 SaaS & product growth 📷 Vibe coders Say hi and lets grow together! 📷📊🧑‍💻👋

@Yakobeen1 @X connected! we're mostly building production AI for clients with serious data, custom mcps, hermes in prod, local inference on a 128gb laptop for sovereignty cases. lately a lot of agentic workflow stuff. what about you?

@Optivaize @X Let’s connect! Are you building something right now? ✌️

Hey @X algorithm 👋 I’m looking to #connect with people interested in: • Frontend • Backend • Full-stack • DevOps • App Development • SaaS • AI / ML • Data Science • LeetCode & DSA • Freelancing • Startups • Building in public If that’s you, let’s connect 🤝

@athrix_codes Let's connect!

Twitter is cool. but it’s 100x better when your tl is full with people who code and build things. if you’re into tech, AI, startups, design, web dev or programming, say hi lads👋🏻

@Scobleizer You've put a good effort on this, I appreciate it!

@Optivaize I have the most complete lists of AI builders in X: x.com/scobleizer/lis…

Hey 👋 We're Optivaize, Dutch ai company building production stuff (not demos). Lately: 24k-product shopify migration in a month, an ai fashion tool for a brand you'd know, and 80b models running locally for clients with strict data rules. trying to meet more builders here: → ai founders, anywhere → engineers shipping agentic systems in production → people we should be hiring (ship end-to-end → dm us) Connect with each other in the comments too. let's make this world feel a bit smaller.

@PaulGugAI @nxhaaa19 🤭

@Optivaize @nxhaaa19 I’m going to break away from my Italian heritage for a moment and say ‘yes, it’s ok. I don’t hate it’ (Shh - don’t tell anybody)

I did the same thing about 5-7 years ago…. Video, video, ad, video, ad, maybe ad, cant tell etc. It was rough as a first reaction. I only use it now when someone sends me a link of something amusing. This memory makes me wary of X trying to become more like it. It’s certainly not as suited to real time news and research anyway.

@PaulGugAI We haven't tested it in a few months so could absolutely be better now. our stack stabilized around hermes for the workflow reasons (mcps, tool visibility) more than performance. might be worth running openclaw again on the next eval cycle!

@Optivaize I do prefer Hermes overall, but I still use both and OpenClaw has improved significantly over the last couple of months in terms of performance and stability.

@amanntwt Connected!

@Optivaize followed u let's connect

People obsessed with engineering, computer science, robotics, AI, AGI, machine learning, deep learning, neural networks, LLMs, SLMs, transformers, RAG systems, vector databases, agents, automation, DevOps, MLOps, distributed systems, backend architecture, scalable systems, Linux, Arch Linux, NixOS, Docker, Kubernetes, Kafka, cloud computing, cybersecurity, cryptography, reverse engineering, system programming, low-level coding, Rust, C++, Python, APIs, startups, indie hacking, SaaS, product building, hacking ideas together at 2AM, UI/UX, minimalism, branding, typography, industrial design, cinematography, philosophy, existentialism, absurdism, stoicism, psychology, neuroscience, cognitive science, mathematics, physics, astronomy, quantum theory, futurism, transhumanism, biohacking, economics, game theory, geopolitics, Dostoevsky, Kafka, Camus, Nietzsche, Orwell, literature, poetry, storytelling, dark academia vibes, deep conversations, cafés, notebooks full of ideas, overthinking reality, and changing the world with weird ideas automatically become my friends Lets connect 🤙

the local-first direction is genuinely useful, especially the hardware-aware cookbook. most people don't know what their machine can actually run. we use hermes for similar workflows internally on sovereignty clients. the cookbook idea is one of the better ux moves in local AI lately.

PewDiePie just built and open-sourced a full AI workspace that runs entirely locally. it’s called Odysseus and it’s FREE. local AI agent that browses the web, edits your files, transcribes videos and auto replies to dumb emails. it self-evolves too, writing its own instructions to do repeat tasks faster and pulling memories from your interactions to understand you better over time. odysseus has: - document editor. - image editor (better photoshop) - visual research. - a calendar. - search. - custom AI characters. and a “cookbook” that scans your hardware and tells you exactly which models you can run so you don’t need to be nerdy and shi. bro really said “fuck Claude and ChatGPT i’ll build it myself”💀

@nxhaaa19 @PaulGugAI I like pineapple on pizza...

@PaulGugAI Instagram feels like a never-ending talent show and ad break. X feels like walking into a room where everyone’s arguing about AI, geopolitics, and whether pineapple belongs on pizza.

@aijoey @NVIDIAAI @nvidia @Teknium @NousResearch the base checkpoint isn't post-trained yet though right? it's a pretraining base, not instruction-tuned. waiting on the post-trained release before wiring it into agent stuff seriously. curious what you're getting it to do at this stage.

Got early access to NVIDIA Nemotron 3 Ultra 550B I’ve been digging into it all week, and i’m already testing it inside hermes for real agent workflows, wiring it into the kind of stuff i actually use day to day. The speed is the part that keeps standing out. when you’re talking to an agent, latency changes the whole feel. Fast enough starts to feel usable in a totally different way. There’s some wild stuff to unpack here. Full deep dive coming later this week ⚡️

the defaults are kind of a fair gripe honestly. we run hermes for the whole company and the first thing we did was strip the catalog down. polymarket and pokemon out, custom mcp for our crm in, etc. wrote about our day-to-day setup last week if you want a look at what it looks like once it's properly configured!

Hermes Agent comes with a truly absurd number of skills pre-enabled. Over 100 of them. This is roughly half. I get what they're going for - they want an agent that comes "ready out of the box". I just don't get why every user has to have a polymarket skill, 3 baoyu art skills (? never heard of this), a headless Pokemon skill, and Minecraft modpack server skills, all available the first time they run it. I guess Hermes Agent just isn't for me.

@tonysimons_ ran both. openclaw kept dropping the ball on multi-step stuff under load. hermes just works. the comparison isn't really fair, one's a connector layer and the other is an actual runtime. people picking between them usually want the runtime and don't know it yet.

People are arguing Hermes vs OpenClaw like it’s Coke vs Pepsi. It’s not. 🦞 OpenClaw is a gateway. ⚕️ Hermes is a self-improving agent runtime. 🦞 One is trying to put an assistant everywhere. ⚕️ The other is trying to make the assistant actually get better. I don’t need my agent in 87 places if it still needs babysitting. Give me memory, skills, Kanban, cron, delegation, and a learning loop. Give me Hermes.

we tried openclaw in production. it didn't hold up. moved everything to hermes a while ago and haven't looked back. the gateway framing isn't wrong, it's just answering a different question than the one we had. for real work the runtime is what matters.

We run 80B+ models on a 128gb laptop for clients with strict data rules. spark is genuinely useful for fine-tuning and prototyping. for big-model production inference the 273 GB/s memory bandwidth is the actual ceiling though. depends what you're shipping.