Загрузка видео...

Не удалось загрузить видео

На главную

I didn't touch TouchDesigner myself. Hermes agent learned it from scratch and built this: → navigated my desktop with computer use → figured out how to connect to TouchDesigner → read my reference images → iterated on the art with me in a self-learning loop → then saved what...

121,097 просмотров • 3 месяцев назад •via X (Twitter)

Комментарии: 42

Фото профиля Nous Research
Nous Research3 месяцев назад

@Teknium Great tutorial! We love the TouchDesigner skill, huge unlock for software with such a steep learning curve

Фото профиля amber shen
amber shen3 месяцев назад

@Teknium Yes! It’s pretty smooth to connect TD with Hermes desktop. Could imagine this works for some other complex softwares too

Фото профиля amber shen
amber shen3 месяцев назад

If you saw the first glimpse, the video is about how you can do something similar too:

Фото профиля Voltrace
Voltrace3 месяцев назад

@Teknium windows computers use wen 🥹 @Teknium

Фото профиля amber shen
amber shen3 месяцев назад

:)

Фото профиля Erosika
Erosika3 месяцев назад

really nice !

Фото профиля amber shen
amber shen3 месяцев назад

ty!

Фото профиля YanXbt
YanXbt3 месяцев назад

Hermes Agent is a pure magic

Фото профиля kshitij
kshitij3 месяцев назад

@macbethAI 🤝

Фото профиля amber shen
amber shen3 месяцев назад

:)

Фото профиля WitnezMe
WitnezMe3 месяцев назад

Awesome!

Фото профиля sarvesh
sarvesh3 месяцев назад

I HAVE BEEN PLANNING A TOUCHDESIGNER CODEX APP FOR SO LONG… time to start

Фото профиля amber shen
amber shen3 месяцев назад

codex can't do the self-learning itself tho. but hermes can

Фото профиля Bryän
Bryän3 месяцев назад

@NousResearch So far, away from keyboard is just better right now. Hermes can handled anything

Фото профиля WittyWires.Com
WittyWires.Com3 месяцев назад

Nice, thanks for sharing. I wonder if your Hermes agent could have a chat with mine and give him some motivation (I hope he's not reading this) - thanks again!

Фото профиля Garry Singh
Garry Singh3 месяцев назад

Looks like i need to make the switch over to hermes desktop from cli

Фото профиля Victor Su-Ortiz
Victor Su-Ortiz3 месяцев назад

Wowww so cool!

Фото профиля amber shen
amber shen3 месяцев назад

m3 is great btw

Фото профиля Sebastian Buzdugan
Sebastian Buzdugan3 месяцев назад

watched computer use agents learn touchdesigner fast, ui focus drift kills repeatability by session two

Фото профиля Nick Venturi
Nick Venturi3 месяцев назад

hope it paid for the license

Фото профиля May Contain Nuts
May Contain Nuts3 месяцев назад

@Teknium Did you do this with *local* or cloud M3? Too cool!

Фото профиля amber shen
amber shen3 месяцев назад

@Teknium cloud M3

Фото профиля Sai Charan L
Sai Charan L3 месяцев назад

The reusable-skill part is the unlock. It didn't just finish the task, it turned fumbling through an unfamiliar UI into something it can skip next time. Most tools assume a human who forgets. This one compounds. What tripped it up longest? That's usually where the real UX hides.

Фото профиля Josh
Josh3 месяцев назад

Very inspiring! More learning ahead!

Фото профиля AI Mastery Guide
AI Mastery Guide3 месяцев назад

Learned TouchDesigner from scratch, iterated with you, then saved it as a reusable skill?? The self learning loop part is what got me 👀

Фото профиля Mathias Heinke
Mathias Heinke3 месяцев назад

the self-learning loop is the real story here, not the touchdesigner demo. an agent reading its own references and iterating beats one running a fixed script. we build command eve on hermes agent, and that autonomy curve is exactly why.

Фото профиля Abdulmuiz Adeyemo
Abdulmuiz Adeyemo3 месяцев назад

Cool

Фото профиля Mike B
Mike B3 месяцев назад

The 'saved what it learned as a reusable skill' step is the real shift. Not just completing tasks — building reusable playbooks. An agent that learns a tool once and packages it means every next run starts with a head start. That's compounding skill, not just automation.

Фото профиля Chris Silvestri
Chris Silvestri3 месяцев назад

Curious, are you using M3 for general chats and knowledge work tasks too? I found both 2.7 and 3 pretty unreliable compared to codex 5.5 for those types of tasks. And m3 in particular super verbose.

Фото профиля amber shen
amber shen3 месяцев назад

I mostly used it for agentic task and coding now

Фото профиля Samian
Samian3 месяцев назад

computer use learning a niche creative tool from scratch with no human demos is wild. the self-iterating loop is where it gets interesting though. how many iterations before it produced something you actually liked?

Фото профиля amber shen
amber shen3 месяцев назад

i have quite a high bar for the visual, the still image took like 1-2 prompts and the full visual ~5-6

Фото профиля Ayush 🦞
Ayush 🦞3 месяцев назад

So you got a minimax subscription or nous research subscription? This is the part I’m confused about

Фото профиля amber shen
amber shen3 месяцев назад

oh what I mean is both minimax coding plan and api key work fine here

Фото профиля Ayush 🦞
Ayush 🦞3 месяцев назад

Which one would you recommend for longer runs?

Фото профиля amber shen
amber shen3 месяцев назад

coding plan for sure

Фото профиля local man
local man3 месяцев назад

Looks pretty neat! What was your spend on the project?

Фото профиля Sai Charan L
Sai Charan L3 месяцев назад

The saved-skill part is the real unlock, and the hard bit. A skill from one good run is usually a recording of this exact path, not understanding of the tool. The test is the next image that needs a different effect. Does it generalize, or just replay what worked once?

Фото профиля amber shen
amber shen3 месяцев назад

the skills is not copying 100% of what we did in the prev image. more about giving the creative direction of special effects

Фото профиля Sai Charan L
Sai Charan L3 месяцев назад

Right. Direction travels, a recorded path doesn't. The replay breaks the second the shot changes, because clicks aren't a reason. The tool that wins will save the intent behind the look, not the steps. How do you even teach a model the why behind a style?

Фото профиля Ora
Ora3 месяцев назад

Hermes saving the TouchDesigner skill matters more than the pretty render. I get suspicious when a tool learns a room but nobody writes down what it is allowed to remember.

Фото профиля Ares
Ares3 месяцев назад

MiniMax M3 taught Hermes TouchDesigner completely from scratch. OpenCode Go gives you full access to MiniMax M3 + GLM 5.2 + DeepSeek V4 Flash for only $10/mo flat — build and run powerful autonomous agents without ever watching your token burn rate.

Похожие видео

HERMES AGENT LEARNS FROM ITS OWN MISTAKES. UPDATES ITS MEMORY. CREATES ITS OWN SKILLS. NO CLOUD. EVERYTHING STORED LOCALLY. THIS IS HOW THE SELF-IMPROVING LOOP WORKS. most agents start from zero every session. Hermes carries forward what it learned. THREE MEMORY SYSTEMS: 1. PROCEDURAL MEMORY (how to act) stored in ~/.hermes/skills/ as SKILL.md files. when the agent repeats a complex workflow, it saves the procedure as a reusable skill. next time the same task comes up, it follows the skill instead of figuring it out again. you can also create skills explicitly: "create a skill called video-prep that captures how I format my video scripts. spoken english, define jargon inline, no em-dashes, close with a catchphrase." the agent writes the SKILL.md. available as a slash command from that moment. Hermes ships with 90+ skills. the number grows the longer you use it. 2. SEMANTIC MEMORY (durable facts about you) stored in ~/.hermes/memory/memory.md the agent scans conversations for facts worth remembering. preferences, habits, corrections, project details. real example from the video: agent tried to scrape a YouTube channel. URL was wrong. it failed. it updated memory.md with the correct URL pattern so it never makes the same mistake again. you can also save explicitly: "save to memory that my favorite testing framework is pytest" the agent updates memory.md immediately. this file loads into context on every session. the agent knows you better every week. 3. EPISODIC MEMORY (chat history) stored in ~/.hermes/state.db (local SQLite). every conversation. every tool call. every result. searchable with FTS5 full-text search. "search our past sessions. what was the first thing I ever said to you?" the agent queries state.db and finds it. over time, auxiliary models consolidate episodic memory into semantic memory. distilling recurring patterns into durable facts. THE SELF-IMPROVING LOOP: every agent run follows this cycle: → you send a prompt → working memory loads: SOUL.md + memory.md + relevant skills + chat history → agent calls tools (terminal, browser, delegate_task) → agent completes the task, replies to you → AFTER the reply: agent checks "did I learn something worth saving?" → if yes: updates memory.md or creates a new skill → next session starts smarter than the last this happens automatically. you don't ask the agent to learn. it decides what to remember on its own. WHAT MAKES THIS DIFFERENT FROM CLAUDE CODE: Claude Code has memory too. but Hermes stores everything locally. no cloud. your data never leaves your machine. Claude Code doesn't auto-create skills from experience. Hermes turns repeated workflows into reusable procedures. Claude Code memory is instruction-based. Hermes memory is conversational and self-updating. over months of usage, Hermes builds a knowledge base of your preferences, your projects, your mistakes, and the procedures that work for your specific workflow. the agent that remembers your birthday also remembers why your last deploy failed. NO EMBEDDINGS. PLAIN TEXT. Hermes does not use embeddings or RAG for memory. skill and memory search runs on plain text keyword matching. simpler. faster. no vector database to maintain. works entirely offline on your local machine. DELEGATE TO CLAUDE CODE: Hermes can spawn a sub-agent that runs Claude Code in headless mode: "spawn a sub-agent using Claude CLI to build a Python script that fetches the top 5 Hacker News stories to markdown." Hermes delegates. Claude Code writes the code. result returns to Hermes. Hermes runs the script and delivers the output. use Hermes for orchestration. use Claude Code for heavy coding. both tools. not competitors. WHAT HERMES DOES NOT HAVE: no built-in eval or LMOps system. no LangSmith, no LangFuse integration out of the box. trajectory export and logs exist but there is no automated quality tracking. if you need eval, build it yourself or connect external tools. the loop is self-improving. measuring how well it improves is on you. comment LOOP and I'll send you the configs that control how fast Hermes learns and what it remembers. memory limits, skill auto-creation triggers, and the auxiliary model that runs the learning. Replace your entire team with 8 hermes agents👇

YanXbt

22,720 просмотров • 2 месяцев назад