halr9000

I am coining here -- Hal's Law of #AI Safety Alignment, which applies broadly to #GenAI, #AGI, and #ASI. "The world's supply of data scientists and security researchers who focus on safety alignment will always be vastly exceeded by trolls."

@cocktailpeanut @PierrunoYT I should dip into the code and add some logging. Lot of people don't bother in open source projects but I am a big fan of o11y. (I work for Splunk, so.)

@halr9000 @PierrunoYT try this one pinokio.co/apps/github-co… and what's your vram?

Run DramaBox on your PC with LOW VRAM Now @PierrunoYT's DramaBox gradio launcher ships with LOW VRAM support! Please test on your machine and report if it works (NVIDIA GPU only) Generated this 30 second audio in 50 seconds on A4500. Prompt: > He raps "His palms are sweaty, knees weak, arms are heavy. There's vomit on his sweater already, mom's spaghetti. He's nervous, but on the surface, he looks calm and ready. To drop bombs, but he keeps on forgetting. What he wrote down, the whole crowd goes so loud. He opens his mouth, but the words won't come out. He's chokin', how? Everybody's jokin' now. The clock's run out, time's up, over, blaow. Snap back to reality, ope, there goes gravity.

@cocktailpeanut @PierrunoYT That's the one I installed. Started normal just to compare. It didn't crash but it also didn't finish after an hour so I killed it . Started it in low vram mode and I was hoping to see it finished in a reasonable time but it didn't. 5060 12GB

@hgruenhagen @openclaw I evaluated a handful of options but nothing seemed to be worth switching and went back to default oai which is just cheap and good enough, given I don't have a GPU on the box I'm running claw. Open to hearing about CPU friendly local options though.

Which embedding models are you using for memory-core in @openclaw ? Mostly curious about retrieval quality and latency.

open-sourcing a 3D gen toolkit for Claude Code input image → environment, meshes, physics, lighting, & audio

WanGP 11.41 to 11.66, Endless flow of goodies: - HiDreamO1 model Image Editor / generator quite good for generating text and for identify transfer - TTS Models Omnivoice & and the lively Scenema Audio - FlashSVR High Quality spatial upsampler is natively supported and works with as low as 6GB of VRAM - Use the Process Full Video plugin to upsample, turn to HDR, recolorize, clean, ... a video of any duration (hint: hours !) - With Vista4D you can replay a video from a different (moving) camera angle - Early Mac / MPS support - One click installation scripts - Lots of new LTX2 knobs to give you more control

@cocktailpeanut ooh that's nice checking out now. I was trying to make an app for a similar recent model release: github.com/OpenSenseNova/… Failed w/ OOM. Found that someone distilled it & supposedly running in comfyui but not working for me yet. Next attempt will start w/ github.com/smthemex/Comfy…

Run HiDream-O1-Image locally with 10GB VRAM HiDream-o1 is a single transformer model (no vae, no separate text encoder) that can reason (like gpt-image-2). I made a 1-click launcher to run the official Web UI, but with 10GB VRAM. Make whatever you want, as much as you want.

@karlprosser Agreed, OP has a shallow take.

To be expected. Transitions are bumpy & uneven , in some ways faster and some ways slower than expected. Widespread adoption often trails capabilities by decades. The “automobile revolution” took 20+ years to get 10% market share of transportation miles vs horses. Or about 5 years from mass consumer products like the model-T. AI, outside of simple chatbot interaction , AI is far from the start of being a mass consumer generic commodity. The place where it’s interesting only has bleeding edge adoption. But it’s probably still faster than the PC adoption from the point of the Apple 1, or the web based business app/e-commerce from the point of residential dialup internet access.

@ibuildthecloud hey you tryna start-something?! get a look at this guy @jsnover @BrucePayette ! ;D

Wow, if you thought insulting the rust community was bad, try powershell.

@watawatabou Apparently the owner decided to block the stairs with furniture :)

@halr9000 But the stairs!

@watawatabou Having fun today using your procgen apps as inspiration for image gen. Here's sigil

😆😆

localhost was right there but he chose the cloud

@openclaw What are people doing with paired nodes? I've not played with that side of the project.

Move files through paired nodes without pretending stdout is a file system. file_fetch, dir_fetch, dir_list, and file_write are policy-gated, default-deny, and capped at 16 MB per hop. docs.openclaw.ai/plugins/refere…

OpenClaw 2026.5.3 🦞 📁 File transfer for paired nodes 🧭 /steer + /side for live agent control 🔌 Plugin installs/updates hardened 🛠️ Channel + upgrade fixes Big release, fewer paper cuts. github.com/openclaw/openc…

Just posted Procgen 9000 #comfyui extension! This gives you a quick way base image for image edit models, example below. github.com/halr9000/procg…

Super happy with this #comfyui extension I'm working on called Procgen 9000. It will procedurally generated images which make for EXCELLENT input to edit models. First up: landscapes!

@cocktailpeanut s/,/--/g

🚨R.I.P {{company A}} Someone just launched {{product B that launched 6 months ago}} and it is wild. It isn't just {{bullshit A}}, it is a whole {{bullshit B}} Here are the top 10 prompts: 🧵

@cocktailpeanut 🤯