home.social

#hermesagent — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #hermesagent, aggregated by home.social.

  1. I finally gave in to the hype and I have to say it seems much more polished than but even with a 5080 (what I am my wallet understood do be a capable gpu) the 64k context limit is brutal. This has led to a whole rabbit hole about how kv cache works and models reserve memory… all to just throw in an api key and give up (again) on having a local llm. At least the 5080 makes my games look pretty lol

  2. Grok Access Simplifies Within Hermes Agent Framework

    Hermes Agent now works with Grok subscriptions without needing separate API keys. This makes AI tasks simpler for users starting May 2026.

    #HermesAgent, #GrokAI, #AISimplified, #NousResearch, #AgentFramework

    newsletter.tf/grok-access-simp

  3. Users with Grok subscriptions can now use the Hermes Agent without extra API keys, making AI tasks much simpler. This is a big change for how AI agents work.

    #HermesAgent, #GrokAI, #AISimplified, #NousResearch, #AgentFramework
    newsletter.tf/grok-access-simp

  4. Local AI Emergence: Hermes Agent Navigates New Frontiers

    Hermes Agent lets you run AI models like Gemma 4 and Qwen 3.5 on your own computer. Find out how it works and why it matters for AI users.

    #LocalAI, #HermesAgent, #LLM, #OpenSourceAI, #TechNews

    newsletter.tf/run-ai-locally-h

  5. You can now run advanced AI models on your computer with Hermes Agent. This is a big change from needing expensive cloud services.

    #LocalAI, #HermesAgent, #LLM, #OpenSourceAI, #TechNews
    newsletter.tf/run-ai-locally-h

  6. There was this idea that I read about some time ago. I don't remember where. Let's find that book so that I can do some more research on it:

    hermes - scrape my public library account, download my reading history, store it in a obsidian file. Search through my reading history to locate book(s) that contain this idea <insert idea>

    Done.

    #hermesagent #genai #reading #library #playwright #seleniumbase #CDP #obsidian

  7. RT @atomicbot_ai: Hermes-Agent im Vergleich zu OpenClaw unter Verwendung des lokalen Qwen 35B-Modells. Wir beauftragten die Agenten, die GitHub-Sterne-Historie für beide Tools zu scrapen, die Ursachen für die Wachstumsspitzen zu ermitteln und ein Live-Dashboard im Browser zu erstellen. MacBook Pro M5 Max mit 64 GB RAM. OpenClaw: 203k Tokens, 12 Minuten 01 Sekunde – schrieb ein Bash-Skript. Hermes: 257k Tokens, 33 Minuten 01 Sekunde – schrieb eine SKILL.md. OpenClaw nutzte die GitHub-API, erhielt jedoch abgeschnittene Antworten; es paginierte durch die Mitwirkenden, lud die JSON-Daten der Sterne-Historie herunter, fand einen Sicherheitsvorfall in der Geschichte von OpenClaw, lud SVGs herunter, reparierte beschädigtes HTML durch Beschneiden und schrieb es sauber neu. Hermes führte parallele Tool-Aufrufe über GitHub-API, Websuche und Browser aus. Es traf auf Google-Ratenbegrenzungen und wechselte automatisch zu DuckDuckGo. Es lud Artikelinhalte herunter, kartierte virale Momente und baute dann das Dashboard. Beide lieferten ein Live-Dashboard mit Sternwachstumsdiagrammen und Anmerkungen zu den Wachstumsspitzen. Video.

    mehr auf Arint.info

    #GitHubAPI #HermesAgent #LiveDashboard #OpenClaw #Qwen35B #WebScraping #arint_info

    https://x.com/atomicbot_ai/status/2055405319455715373#m

  8. Higgsfield AI veröffentlicht den autonomen Agenten »Supercomputer« zur End-to-End-Videoproduktion.

    Die Hermes-Logik-Engine orchestriert Aufgaben über 40 integrierte Tools und wählt eigenständig externe KI-Modelle wie Sora 2 oder Kling 3.0. Erste Berichte dokumentieren hohe Kosten von rund 700 Credits für fünf Sekunden Video.

    #HiggsfieldAI #Supercomputer #HermesAgent #ContentCreation #AIGeneratedImage

    all-ai.de/news/news26top/super

  9. Higgsfield AI veröffentlicht den autonomen Agenten »Supercomputer« zur End-to-End-Videoproduktion.

    Die Hermes-Logik-Engine orchestriert Aufgaben über 40 integrierte Tools und wählt eigenständig externe KI-Modelle wie Sora 2 oder Kling 3.0. Erste Berichte dokumentieren hohe Kosten von rund 700 Credits für fünf Sekunden Video.

    #HiggsfieldAI #Supercomputer #HermesAgent #ContentCreation #AIGeneratedImage

    all-ai.de/news/news26top/super

  10. Higgsfield AI veröffentlicht den autonomen Agenten »Supercomputer« zur End-to-End-Videoproduktion.

    Die Hermes-Logik-Engine orchestriert Aufgaben über 40 integrierte Tools und wählt eigenständig externe KI-Modelle wie Sora 2 oder Kling 3.0. Erste Berichte dokumentieren hohe Kosten von rund 700 Credits für fünf Sekunden Video.

    #HiggsfieldAI #Supercomputer #HermesAgent #ContentCreation #AIGeneratedImage

    all-ai.de/news/news26top/super

  11. First in a series of live-streams evaluating and other tools for OSS coding on an existing project by a naive fool with a vanishingly small token budget.

    youtu.be/c7CgYjAzNFM?si=83DlxJ

  12. Llevo ya un par de semanas dandole una oportunidad a #hermesAgent y de momento parece que cumple las expectativas. Ademas lo de correr en local modelos de IA lo lleva bien, asi que parece que Anthropic, Google, OpenAI y similares no verán un duro mas por mi parte...

  13. Hermes Desktop is what you use if you like the idea of Hermes Agent but not the setup.

    Normally, you’d install it through the terminal, deal with configs, APIs, and hope nothing breaks. Here, you just install an app and open it. It sets things up, asks what provider you want to use, and drops you into a working interface.

    You still get all the agent stuff, tools, memory, integrations, but without the usual issues.
    firethering.com/hermes-desktop
    #ai #technews #hermesagent #opensource #genai #hermes

  14. There is no reliable and user-friendly way to isolate AI agents on macOS other than separate hardware. Every emerging sandboxing tool is a bundle of compromises that falls apart under real workflows.

    I’m still exploring but what I’ve seen is that it’s either back to the Linux From Scratch times, pinky-promises built on deprecated dependencies or good ideas that only work for very simple demos.

    #aiagent #security #agentsIsolation #agentsSandboxing #microVM #claude #opencode #hermesAgent

  15. Mein lokaler Hermes Agent, powered by nousresearch/hermes-4-70b

    So schnell werden die Maschinen hier gar nix übernehmen.

    #ai #agenticai #hermesagent #nousresearch

  16. AI Should Raise the Bar, Not Lower It

    AI can speed up the right kind of work, but it also has a talent for scaling whatever standards you bring to it. Used well, it helps careful engineers move faster; used badly, it turns half-baked thinking into a much more efficient mess.

    jsynowiec.github.io/posts/ai-s

    #AI #generativeAI #AIEngineering #AIAssistant #OpenClaw #Claude #musing #genai #HermesAgent #softwareEngineering #CriticalThinking #DeepWork

  17. AI Should Raise the Bar, Not Lower It

    AI can speed up the right kind of work, but it also has a talent for scaling whatever standards you bring to it. Used well, it helps careful engineers move faster; used badly, it turns half-baked thinking into a much more efficient mess.

    jsynowiec.github.io/posts/ai-s

    #AI #generativeAI #AIEngineering #AIAssistant #OpenClaw #Claude #musing #genai #HermesAgent #softwareEngineering #CriticalThinking #DeepWork

  18. AI Should Raise the Bar, Not Lower It

    AI can speed up the right kind of work, but it also has a talent for scaling whatever standards you bring to it. Used well, it helps careful engineers move faster; used badly, it turns half-baked thinking into a much more efficient mess.

    jsynowiec.github.io/posts/ai-s

    #AI #generativeAI #AIEngineering #AIAssistant #OpenClaw #Claude #musing #genai #HermesAgent #softwareEngineering #CriticalThinking #DeepWork

  19. AI Should Raise the Bar, Not Lower It

    AI can speed up the right kind of work, but it also has a talent for scaling whatever standards you bring to it. Used well, it helps careful engineers move faster; used badly, it turns half-baked thinking into a much more efficient mess.

    jsynowiec.github.io/posts/ai-s

    #AI #generativeAI #AIEngineering #AIAssistant #OpenClaw #Claude #musing #genai #HermesAgent #softwareEngineering #CriticalThinking #DeepWork

  20. AI Should Raise the Bar, Not Lower It

    AI can speed up the right kind of work, but it also has a talent for scaling whatever standards you bring to it. Used well, it helps careful engineers move faster; used badly, it turns half-baked thinking into a much more efficient mess.

    jsynowiec.github.io/posts/ai-s

    #AI #generativeAI #AIEngineering #AIAssistant #OpenClaw #Claude #musing #genai #HermesAgent #softwareEngineering #CriticalThinking #DeepWork