boringasdead retweeted
Une vraie saloperie la macronie.
Vivement 2027 que cette gangrène dégage à tout jamais..
J'en peux plus de ces FDP..
boringasdead retweeted
Après Elon Musk, j’ai rassemblé toutes les interviews de Jensen Huang depuis 2003 sur un seul site.
283 vidéos. 241 heures. 1 867 citations. Et 14 de ses prédictions vérifiées une par une.
Tout se regarde directement sur le site, en 8 langues.
silicon-valley.tech
boringasdead retweeted
Avec le débat de ce soir, Jean-Luc Mélenchon confirme qu’il est le seul candidat capable d’éradiquer la misère dans ce pays et de battre Marine Le Pen, qui a confirmé son hostilité totale aux associations et son projet de mise au pas du monde associatif français.
boringasdead retweeted
finally got a mac??
here's everything you need to install
- homebrew
- raycast
- ghostty
- claude
- t3 code
- opencode
- grokbot
- cursor / zed / athas
- wonder / paper
- aside / dia / ego lite / helium
- notion / obsidian
- recordly / shotbase
- gatherOS / pond
- rex multiplexer
- vorssaint
- rectangle
- droppy
- tilde
- mac mouse fix
- stremio
Best uncensored AI models you can run on your own computer right now
No API.
No subscription.
No one reading your chats.
Organized by how much GPU memory (VRAM) you have:
8GB VRAM (most gaming laptops)
→ Dolphin 3.0 8B
Built on Llama 3.1. Trained on filtered data, so it holds its quality. Good at coding, math and chat. 128K context.
→ Hermes 3 8B
Not "uncensored," it's neutral. It does what your system prompt tells it. The least quality loss of anything on this list.
→ Qwen3 Abliterated 8B
A newer base model with the refusal reflex removed.
12 to 16GB VRAM
→ Gemma 4 Abliterated 12B
The most downloaded abliterated line. Reads images too. 128K context.
→ Qwen3 Abliterated 14B
The sweet spot for 16GB cards.
→ gpt oss Abliterated 20B
OpenAI's open model with tool use and thinking.
24GB VRAM (RTX 3090 / 4090 class)
→ Qwen 3.8 Abliterated 27B
The newest base. 256K context.
→ Gemma 4 Abliterated 26B
Big context, reads images.
48GB+ VRAM
→ Hermes 3 70B
→ Llama 3.3 Abliterated 70B
→ DeepSeek R1 Abliterated 70B for reasoning
Here's what nobody tells you:
"Abliterated" means someone found the refusal pattern inside the weights and removed it.
It's not free.
These models get measurably sloppier at following instructions and multi step reasoning.
Dolphin and Hermes are fine tunes instead, so they hold up better.
The smart setup:
Keep a normal model for real work.
Keep an uncensored one for the jobs it refuses on.
Switch by task.
What people actually use them for:
Fiction with dark themes.
Security research and red teaming.
Medical and legal questions answered straight.
Total privacy, because nothing leaves your machine.
How to run any of them:
1. Install Ollama
2. Search the model name on ollama.com
3. Copy the run command
4. Done
Check each model's license before using it in anything commercial.
Download sizes and exact commands in the replies 👇
boringasdead retweeted
Ils créent une génération qui vont détester les flics à un niveau jamais vu



































