Open Source it all https://xitter.cf/t.co/Nc3P68V5Em
The Genesis Block
Joined December 2010
- Tweets11.8K
- Following2.2K
- Followers6K
- Likes34.6K
Recently finished a bike ride and saw kittens scattering into middle the forest.
Wondering WTF was going on, the little one approached me, meowing with conviction. I found a box. Some twat abandoned them for dead.
Spent the next few days/nights searching for all of them….
They’re getting on famously with Dave Growl and Izzie Rascal, 2x 12 year old sausage dawgs.
Both parties very curious of each other but finally multi hugging and sharing the love…
Raptoooor retweeted
Bitfest isn't only looking for people with a polished keynote.
If you're building, teaching, organising, making, testing or explaining something useful around Bitcoin, we want to hear from you.
Workshops, demos and practical contributions matter too.
Am I completely tripping or did anyone else’s iPhone automatically play the Interstellar music automatically when looking at their photo of the solar eclipse?!
ChaosEngineAI v0.9.4
🆕 4 new frontier families in Discover:
• DeepSeek V4 (685B MoE, 128K ctx)
• GLM-5 (130B MoE, 128K ctx)
• Gemma 4 E2B + 31B — fully multimodal, vision built-in
• MiniMax M2.7 (256 experts, 200K ctx)
⚡ Chat is faster:
• Persistent MLX prompt cache — turn 2+ is ~5.6× quicker
• Cross-turn KV reuse on llama.cpp too
• DRY + XTC samplers in the Advanced panel
• Backend cold start: 2.6s → 0.85s
🔓 More open:
• Local API endpoint — any OpenAI-compatible tool just works
• One-click RAG — install the embedding stack from Setup, attach docs to chat
• Run any Hugging Face model — paste a repo, hit load, done
📦 Deps: turboquant-mlx-full 0.8.0 (Mamba/hybrid arch), vLLM 0.23.0, mlx-vlm 0.6.3, Tauri 2.11.2
ChaosEngineAI v0.9.2: MTP speculative decoding everywhere ⚡️
-MTPLX (Apple Silicon)
-GGUF MTP via llama.cpp
Plus ChaosEngineAI-CLI + 8-phase E2E suite, and 1-click installs for Nunchaku, SageAttention, DFlash, vLLM-WSL2.
github.com/cryptopoly/ChaosE…
ChaosEngineAI v0.9.2: MTP speculative decoding everywhere ⚡️
-MTPLX (Apple Silicon)
-GGUF MTP via llama.cpp
Plus ChaosEngineAI-CLI + 8-phase E2E suite, and 1-click installs for Nunchaku, SageAttention, DFlash, vLLM-WSL2.
github.com/cryptopoly/ChaosE…
ChaosEngineAI v0.9.0 🌐
Full UI now in 10 languages:
🇬🇧 🇨🇳 🇹🇼 🇯🇵 🇩🇪 🇷🇺 🇰🇷 🇫🇷 🇪🇸 🇧🇷
20,240 strings translated. Settings → Language, restart-survives.
Plus 25 commits of refactor, audit, and bug fixes since v0.8.
Local LLMs that speak your language
github.com/cryptopoly/ChaosE…
Raptoooor retweeted
ChaosEngineAI v0.7.6:
🧪 HTML Challenge — 2-4 LLMs, same prompt, side-by-side live HTML previews + validation
🔁 Retry failed turns one-click
🔍 UI scale 75/100/125/150%
🎬 Wan 2.1/2.2 on Apple Silicon
🎨 FBCache, live denoise thumbnails, MLX prompt enhancer
github.com/cryptopoly/ChaosE…
Many Improvements in new version v0.7.4
💬 Chat: branching, compare, RAG, MCP, mid-turn model swap
🎨 Image: FBCache, live thumbnails, MLX prompt rewrite
🎬 Video: Wan 2.1/2.2 on Apple Silicon, end-to-end
🖥️ CUDA: SVDQuant + FP8 layerwise
github.com/cryptopoly/ChaosE…
llama.cpp + spec... MTP support merged 👀🔥👀
github.com/ggml-org/llama.cp…
ChaosEngineAI v0.9.0 🌐
Full UI now in 10 languages:
🇬🇧 🇨🇳 🇹🇼 🇯🇵 🇩🇪 🇷🇺 🇰🇷 🇫🇷 🇪🇸 🇧🇷
20,240 strings translated. Settings → Language, restart-survives.
Plus 25 commits of refactor, audit, and bug fixes since v0.8.
Local LLMs that speak your language
github.com/cryptopoly/ChaosE…
ChaosEngineAI v0.7.6:
🧪 HTML Challenge — 2-4 LLMs, same prompt, side-by-side live HTML previews + validation
🔁 Retry failed turns one-click
🔍 UI scale 75/100/125/150%
🎬 Wan 2.1/2.2 on Apple Silicon
🎨 FBCache, live denoise thumbnails, MLX prompt enhancer
github.com/cryptopoly/ChaosE…
Many Improvements in new version v0.7.4
💬 Chat: branching, compare, RAG, MCP, mid-turn model swap
🎨 Image: FBCache, live thumbnails, MLX prompt rewrite
🎬 Video: Wan 2.1/2.2 on Apple Silicon, end-to-end
🖥️ CUDA: SVDQuant + FP8 layerwise
github.com/cryptopoly/ChaosE…
Many Improvements in new version v0.7.4
💬 Chat: branching, compare, RAG, MCP, mid-turn model swap
🎨 Image: FBCache, live thumbnails, MLX prompt rewrite
🎬 Video: Wan 2.1/2.2 on Apple Silicon, end-to-end
🖥️ CUDA: SVDQuant + FP8 layerwise
github.com/cryptopoly/ChaosE…
Local AI Video Generation with
LTX2.2 / LTX2.3 / WAN2.2
ChaosEngineAI 0.7.2
Not quite passing the 'Will Smith eating spaghetti" test but getting there.
Open Source ✅
github.com/cryptopoly/ChaosE…
Local AI Video Generation with
LTX2.2 / LTX2.3 / WAN2.2
ChaosEngineAI 0.7.2
Not quite passing the 'Will Smith eating spaghetti" test but getting there.
Open Source ✅
github.com/cryptopoly/ChaosE…
ChaosEngineAI now featuring more stuff for doing Local LLM things 😅
🗜️ TurboQuant / RotorQuant / TriAttention
🏎️ DFlash / DDTree
🎨 Image generation
🔍 Chat Compare
💽 MLX Conversion
github.com/Cryptopoly/ChaosE…
