Building with Local AI | DGX Spark x2 Health-Tech | EdTech Need more VRAM. Father of 2. https://nitter.cf/t.co/9MPYYPGGFV 🎧 "Health AI Morning Briefing" 🍎 Podcasts & Spotify
Reddit- DGX Spark
Joined November 2008
- Tweets4.9K
- Following1.3K
- Followers743
- Likes8.2K
Pinned Tweet
I’m genuinely blown away by this community that has come together on Twitter around local AI. Supporting each other, with expertise, time, energy, ideas, work, inference (yes) and financially. It’s wild to see the collaboration and support.
@ashxhart - who built Tensorfold, arguably the best local inference engine out there right now, got turned down for a loan to purchase a Mac Studio because his income “wasn’t consistent enough”. In 24 hours, the community came together and crowdfunded it to the tune of $32,000+ USD. Unreal.
We did it! Local AI community is so powerful!
Less than 24 hours to help @ashxhart to get an M5 Ultra 512GB to push TensorFold to the next level!
WE ARE THE OPEN AI! 💪
Inching my way up the @spark_arena leaderboard!!! Got a few more benches coming up. I’m coming for you @DrewBotwinick
Can confirm. This thing is amazing. @NousResearch crushing it out of the gate.
So annoyed I missed this.
I’m already loving the Hermes mobile app. @NousResearch @witcheer @Teknium you guys keep crushing it. Thank you and keep it up. We appreciate it. 🙏
Guys - so many recipes out there for Deepseek 4.1 1017. This is, STILL, by far, the best one I have used. Better than all the @MiaAI_lab recipes. Faster, consistent, no denials. TRUST ME. If you have 2 sparks. This is the one you want.
This is exactly why I love being part of this community. Way to go @SamJWasserman and @u1tra_instinct for building for the community on FOSS - you guys are crushing it and I will be SHOCKED if this doesn’t win the competition.
@ComfyUI #OpenSourceHiggsfield Submission:
Instead of 1 feature, we built the entire Higgsfield platform + way more... but open source.
🎬 Imagination Studio.
An all in one studio with 61+ apps. Video,Music,Audio, Sfx/Foley, Agents, Prompt Assistance, Depth Mapping, Audio Separation, Loras, Custom 3D animatics, Editing.
Collab by Me and @u1tra_instinct
REPO LINK BELOW: 👇
github.com/drowzeys/Imaginat…
Way to go @ComfyUI for not only holding @higgsfield accountable but for bringing the community together to benefit us all and for supporting the community. I remember another certain “open” LLM company that promised the same and hasn’t been held accountable
@higgsfield_ai sent us a cease and desist last night, asking us to shut down our Open Source Higgsfield challenge.
Our read on why they sent it. When the apps people pay for can be rebuilt in the open in two weeks, that’s an uncomfortable thing for a closed platform to watch.
So here’s us reiterating our promise to every builder in the challenge and community, and we’re keeping it:
• The challenge stays open
• Final submissions close tomorrow, Oct 30
• Winners are announced Nov 6, with prizes paid as promised
You showed up for each other. We’re showing up for you.
If we win, we all win!
Comfy out 🫡
James Aita retweeted
@higgsfield_ai sent us a cease and desist last night, asking us to shut down our Open Source Higgsfield challenge.
Our read on why they sent it. When the apps people pay for can be rebuilt in the open in two weeks, that’s an uncomfortable thing for a closed platform to watch.
So here’s us reiterating our promise to every builder in the challenge and community, and we’re keeping it:
• The challenge stays open
• Final submissions close tomorrow, Oct 30
• Winners are announced Nov 6, with prizes paid as promised
You showed up for each other. We’re showing up for you.
If we win, we all win!
Comfy out 🫡
Last night, @higgsfield sent @ComfyUI a cease and desist (attached)
They want the Open Source Higgsfield challenge taken down within 48 hours.
Their argument is that our use of the Higgsfield name and that using their CEO's likeness is an infringement violation (man, the irony). Claiming that it makes the event look like Higgsfield is associated with it.
This couldn't be farther from the truth, which is that Higgsfield made a promise to open source its product and never did.
For the past two weeks, builders have been recreating popular closed apps as open workflows anyone can run.
So we asked our community to deliver where Higgsfield failed: apps in the open, built by the community, for the community. We've made it clear that the challenge is organized by ComfyUI and not affiliated with or endorsed by Higgsfield.
When we started Comfy, we made the community a promise. Our product stays open, free, and yours. We’ve kept it every day since, and one letter isn’t going to change that.
The challenge stays up. Comfy and open source will win!
This is one of my favourite projects yet. If you have a StackChan - check this out, I’ve already pulled Alex’s project and am merging it with my own. Love the animations and DJ modes.
This is my Stackchan Tribute of Nous-Girl, a little desk robot I've been building for fun. She talks, listens, makes faces, dances, plays music, does DJ and rave mode, shows memes, and can even send email.
Repo: github.com/nous-girl-stackch…
I'm not an expert at this, but tried my best to write this guide, it's a work in progress. Will keep adding to it as I have time! Adding more features hopefully everyday.
What you need: an M5Stack StackChan (CoreS3), a USB-C cable and a computer (Windows, Mac or Linux). For memory, intelligence and extra skills, you will need Hermes-Agent from @NousResearch.
The easiest way to start:
Download a beta build with a starter face (robot, boy or girl).
Flash it from your browser or with one command.
On first boot, point it at the official xiaozhi cloud and start talking. No server needed.
Want more privacy and control? Run your own server for the fancy stuff, like DJ mode, memes, email and adding a smarter Hermes Agent brain. Every feature has step-by-step docs.
Heads up: this is beta. I've only tested on my own robot, so things may break. Please report issues on GitHub and I'll help the best I can.
Coming next: simpler builds that need less setup and less upkeep, a version that runs fully from the Nous Research cloud, more languages and local models.
Repo: github.com/nous-girl-stackch…
(StackChan + Hermes Agent Bot)
@M5Stack StackChan Robot
@agentmail Robot EMail
@GradiumAI Voice TTS
@ollama Local LLM
@bot Hermes partner in crime
Just tested this on my own system and it works just like that. FAST AF on a M4Max, and wild what it recognizes in a split second.
Just launched it -- check it out! We use embeddings because they are insanely fast and cheap to "retrain"/refit. For a lot of use cases embeddings are all you need, but the math and experimentation made it challenging.
We put a systemone endpoint in the mix to remove a lot of the hard thinking about the right way to make use of embeddings.
nitter.cf/mmastrac/status/211550…
My @NousResearch Hermes agent running Deepseek uncensored 4.1 successfully rooted, booted and loaded a 2025 FireTv (thought to be unhackable) and loaded @dhh @omarchy TV. God I love local AI.
Model drop! Clover 35B
Meet Clover-35B-A3B-Preview 🍀 the first open-weight production diffusion language model (dLLM) with 35B total parameters capable of 680 tps on a single card at BF16.
Clover exhibits strong performance against both dLLM and AR models and is available today at @huggingface @ModelScope2022 and at our website
